Darren Ang

The Web Scraper’s World of Copyright Exceptions and Contractual Overrides

A PDF version of the article can be found here.


The Web Scraper’s World of Copyright Exceptions and Contractual Overrides

 

Darren Ang*

 

“Web scrapers” are software programs designed to automate the copying of data from the Web, and they can be used for a variety of useful applications. By building upon a technical understanding of how web scrapers operate, this article discusses the two main areas of law which control the use of web scrapers (“web scraping”): copyright law, which provides for certain exceptions that may be relied upon by users of web scrapers to avoid issues of copyright infringement, and contract law, which might bind users of web scrapers under webpages’ terms of service and restrict them from conducting web scraping activities. Following which, this article turns to the growing number of statutory regimes which prohibit the use of contractual terms to exclude the operation of copyright exceptions, and considers the circumstances under which web scrapers may rely on these regimes.

 

I.               Introduction

 

“Web scrapers”,[1] technically defined, are software programs which are “designed to automate the downloading and parsing of the data from the Web”.[2] They operate by interacting with the Web’s communication protocols to extract data from webpages,[3] and they can do so at astonishing speeds. The data extracted from the use of web scrapers (“web scraping”) can then be used for a wide variety of applications.[4]

However, some of these applications are more desirable than others. On one hand, some of the most ground-breaking features of the Web, such as search engines and optimised search results, require the use of web scrapers for their implementation.[5] On the other, web scrapers have facilitated fraudulent acts such as the creation of fake accounts for credit card scams.[6] Additionally, a carelessly-programmed web scraper might cause a website to be overloaded with indiscriminate traffic, causing it to slow down or go offline entirely.[7] These “distributed denial-of-service attacks”, or as they are more commonly known, “DDoS attacks”, attract criminal liability in most jurisdictions.[8] While it is a standard practice for programmers of web scrapers to set a reasonable time delay between requests to the Web,[9] the risks to website owners still remain.

Therefore, for societies to reap the benefits of web scrapers without bearing too much of their risks, entire arsenals of legal controls have been deployed against the use of web scrapers, as well as certain applications created upon web scrapers’ outputs.[10] This article shall discuss the current state of the law regarding two of those controls: copyright law and contract law. Its primary aim is to provide readers with a general map of the complex frameworks covering this area, in a manner that may accommodate the regimes of as many jurisdictions as possible. It does so by drawing examples from multiple jurisdictions and attempting to define common threads.

Following this introduction, Part II sets the foundation of this article by providing a technical overview of how web scrapers can be used to create useful applications. Parts III and IV build upon this technical foundation, and discusses the copyright and contractual considerations of web scraping respectively. Part V discusses the intersection between these two dimensions, found in the prohibitions against using contractual terms to override the operation of statutory copyright exceptions. Finally, Part VI summarises the implications of these legal frameworks for users of web scrapers, and concludes the paper with some final thoughts about the ethics of web scraping.

 

II.             How Web Scraping Works

 

At the outset, a distinction must be drawn between the use of web scrapers themselves and the creation of applications based on web scrapers’ outputs.

In short, a web scraper merely creates copies of webpages’ source codes, which are unlikely to be of much use on their own. However, the source codes may then be processed and analysed in ways that result in useful applications. These two steps often come hand-in-hand, and they shall be discussed in turn below.

 

A.    Step 1: The use of web scrapers in themselves

 

1.     How web scrapers work: A technical overview

 

We begin with web scrapers themselves. Conventionally,[11] most web scrapers are programmed to do a single task: they look up the underlying source codes of webpages and copy the parts of those webpages which are relevant to the programmer.

The source code which all webpages are built upon is known as HyperText Markup Language, or HTML.[12] When we surf the Web on a web browser, the “pages” which we see on our screens are formed through the rendering of webpages’ HTML codes—that is, every word, image, or video which appears on a webpage would have been derived from some expression in HTML code.[13] As a classic example, for a web browser to display a simple webpage titled “Hello”, with the plain text “Hello, world” on it, the following HTML code would have to be rendered:

<html><title>Hello</title><body>Hello, world</body></html>

Webpages are accessed through their unique Uniform Resource Locators, or URLs.[14] By sending a “request” to a particular URL, a web browser may identify the HTML code of the webpage at that URL and render it. The same process is used by web scrapers to copy the HTML codes of webpages—they are designed to send “requests” to webpages and copy the HTML codes associated with them.[15] The programmer determines the parts of the HTML code which are relevant for the application they intend to create, and copy only those parts.[16]

 

2.     The Robots Exclusion Protocol and “robots.txt” files

 

No discussion of web scraping is complete without a brief nod to the Robots Exclusion Protocol (the “REP”).[17] Under this protocol, website owners may include a “robots.txt” file on their website,[18] which tells web scrapers which parts of it they are permitted to scrape based on certain conventions.[19]

Compliance with the REP is strictly voluntary, and programmers of web scrapers are free to ignore websites’ “robots.txt” files if they choose to do so.[20] However, as a matter of etiquette, the general practice is for programmers to ensure that their web scrapers comply with websites’ “robots.txt” files;[21] major search engines such as Google and Bing are programmed in this manner.[22] Google estimates that there are about half a billion websites which rely on this protocol.[23]

 

B.     Step 2: The use of web-scraped HTML code for useful applications

 

Once a web scraper has copied the HTML code of a webpage, this code may then be processed and used for a wide variety of applications. For example, Google’s search engine is powered by the processing of huge amounts of webpages’ HTML codes, which are obtained through web scrapers “run simultaneously by thousands of machines”.[24] Large amounts of HTML code can also be used in conjunction with statistical techniques to yield empirical conclusions,[25] which in turn, can be used for a variety of purposes.

 

III.           Copyright Exceptions Relevant to Web Scraping

 

A.    Why web scrapers might infringe copyrights

 

The most immediate legal concern for users of web scrapers is that their web scraping activities might infringe the copyrights of webpage owners. Computer code, which includes HTML, is explicitly classified as a type of work that may be protected by copyright under the TRIPS Agreement, which currently has 164 state parties.[26] With this in mind, the ability of web scrapers (in themselves) to copy voluminous amounts of HTML code from multiple webpages might immediately raise concerns of copyright infringement in many jurisdictions across the world.[27]

However, not all HTML code is protected by copyright, as copyright protection only extends to “expressions” and not to “ideas”.[28] In many jurisdictions, the line between “ideas” and “expressions” is drawn by the concept of “originality”, which in turn, is a fact-sensitive inquiry that looks towards the creative efforts expended by the author of a work.[29] Therefore, when considering whether web scraping might result in copyright infringement, the real question to be asked is generally whether the HTML code copied by web scrapers is “original” enough to warrant copyright protection.

This turns out to be an extremely difficult question to answer. While only an “extremely low” level of originality is required for a work to be protected by copyright,[30] there are many features of HTML code which make even this low threshold hard to cross: the code might be seen as “factual”, rather than creative, in nature;[31] or it might be seen as being primarily dictated by functional purposes rather than authorial creativity.[32] Ultimately, it is difficult to determine whether web scraping raises copyright issues, as it is difficult to determine whether specific blocks of HTML code may be protected by copyright in the first place.

Nonetheless, the vexed issues surrounding web scraping and copyright infringement can be avoided by turning to the limitations and exceptions found in most copyright statutes across the world (henceforth referred to as the “copyright exceptions”). This article shall discuss two copyright exceptions which are particularly relevant for web scraping: those for text and data mining (“TDM”), as well as the more open-ended “fair use” and “fair dealing” exceptions.

 

B.     Text and data mining exceptions

 

1.     The policy of the TDM exceptions

 

To alleviate the legal uncertainty associated with TDM activities, which may include certain applications of web-scraped HTML code, many jurisdictions have introduced specific copyright exceptions for them.[33]

TDM activities are generally defined as automated or computational analytical techniques used to extract new information from existing data.[34] Crucially, some form of analysis must be done—the underlying data cannot be copied without anything more. A standard example of an act which would not be exempted by the TDM exceptions is the use of an automated tool (such as a web scraper) to collate works into a database, without performing any additional analysis on the works.[35]

In general, the TDM exceptions prevent TDM activities from giving rise to liability for copyright infringement when they are done appropriately. Their shared underlying policy is that TDM activities are beneficial for the economy and society-at-large, and as a result, copyright law should not inhibit them.[36] Conversely, returning to the example from the preceding paragraph, the mere use of automated tools to copy data, without performing any additional analysis on it, would not be consistent with the exceptions’ underlying policy.

 

2.     Common features of TDM exceptions across jurisdictions

 

While the exact scopes of the TDM exceptions adopted across jurisdictions differ, they share some common features, and two of them shall be discussed below.

First, the TDM exceptions are purpose-based. That is, the exceptions can only be relied upon if the relevant TDM activities fall within the spectrum of permissible purposes, which are delineated by the relevant statutory provisions. On the narrower end of the range of TDM exceptions is the UK’s TDM exception, which only allows copies of a work to be made “for the sole purpose of research for a non-commercial purpose”.[37] On the other end, the TDM exception in Singapore’s Copyright Act allows for copies of a work to be made for the purposes of “computational data analysis”, which in turn, is defined broadly and non-exhaustively.[38]

Second, most of the TDM exceptions require the person undertaking TDM activities to have “lawful access” to the works which are being mined.[39] While the exact scope of the term “lawful access” has not been defined within these exceptions, a standard example given by many jurisdictions is having a paid subscription to access a database of works.[40]

 

3.     How the TDM exceptions might apply to cases involving web scraping

 

How, then, do the TDM exceptions apply to web scraping? It bears recalling that web scraping is a common precursor to the analysis of that code,[41] but it does not constitute any form of analysis in and of itself—it only copies the HTML codes of webpages in some form.

Therefore, the act of web scraping without anything more would not satisfy the TDM exceptions, and it is only the use of web scrapers for the purpose of creating applications upon the web-scraped data which may qualify for the exceptions, subject to the relevant requirements being met.

 

C.    Open-textured “fair use” and “fair dealing” exceptions

 

Users of web scrapers might have another string to their bows—the open-textured “fair use” and “fair dealing” exceptions, which are available in more than 40 countries around the world.[42] These, at least on their face,[43] fall on a spectrum of flexibility: at one end are the narrower “fair dealing” exceptions which exhaustively state the purposes which may qualify for them,[44] and at the other are the broader “fair use” exceptions which adopt non-exhaustive factors as guidelines for courts to determine whether the use was fair.[45]

Despite the stark differences in statutory language, the “fair use” and “fair dealing” exceptions share a common theme. They both recognise that copyright is ultimately a balancing act: it must incentivize private individuals to make creative works, while also preserving the public interest in the dissemination of creative works.[46] Following from which, some works which appear to be copyright-infringing might actually promote the underlying goals of copyright under their specific circumstances.[47] The inquiry as to whether a use was “fair” ultimately collapses into a case-by-case inquiry that takes into account all the relevant factors.[48]

These exceptions are generally wide enough to encompass a whole range of activities, and it is unsurprising that they have been raised in some cases involving allegations of copyright infringement caused by web scrapers.[49] Therefore, if a particular jurisdiction does not have a TDM exception, or if an application created upon web-scraped HTML code would not satisfy the requirements of the TDM exception, the “fair use” and “fair dealing” exceptions might still be relied upon to avoid liability for copyright infringement.

 

IV.           Contractual Restrictions to Web Scraping

 

Although users of web scrapers might escape liability for copyright infringement by relying on the exceptions discussed above, they might find themselves facing contractual liability under contracts entered into between them and website owners.

The most common way for this liability to arise is under websites’ terms of service. Additionally, considering the centrality of the REP to web scraping, the possibility of “robots.txt” files being incorporated into websites’ terms of service shall be briefly considered.

 

A.    Websites’ terms of service

 

Many websites contain a page of terms and conditions that purport to govern their use (the “terms of service”). These terms of service might contain specific restrictions against the use of web scrapers,[50] or more general restrictions against the unauthorised use of any content hosted on the website.[51]

Naturally, users of web scrapers might face contractual liability for breaching these terms—this might mean having their access restricted or their account terminated.[52] Where applications are created upon web-scraped data, the creators of those applications might also face tortious liability for inducing the users of their applications to breach websites’ terms of service.[53]

However, the above analysis proceeds on the assumption that the users of web scrapers are bound by websites’ terms of service in the first place. This assumption does not always hold true. In general, the enforceability of websites’ terms of service remains a controversial issue,[54] and it has been argued that the question should be answered with reference to basic principles of contract law.[55] In this regard, it is notable that a two-part test has been developed in US case law to guide the inquiry of whether a website’s terms of service are enforceable against a user:[56]

(1)   “[W]hether the terms were ‘reasonably communicated’ to the user; and

(2)   [W]hether the terms were accepted by the user.”

Determining whether the user of a web scraper is bound by a website’s terms of service is a complicated exercise. However, once this hurdle is crossed, the user of the web scraper would generally be bound by the terms stipulated in the website’s terms of service, which might include restrictions on their web scraping activities.

 

B.     Websites’ “robots.txt” files?

 

Unlike websites’ terms of service, which have been found to have binding force in several cases, the legal status of “robots.txt” files remains entirely uncertain.[57] Nonetheless, as the REP is a highly-established industry standard for programmers of web scrapers, it is conceivable that a case may arise in future where it is argued that a website’s “robots.txt” specifications binds web scrapers under some form of contract. This article shall attempt one such argument.

From a first-principles perspective, it may be possible for the specifications of a “robots.txt” file to be incorporated as terms into the terms of service of a website,[58] due to the status of the REP as a widely-adopted industry standard—it might be reasoned that those who use web scrapers can be taken to subjectively know that they should comply with the REP.[59] However, as the REP is strictly voluntary in nature, this may militate against a finding that the parties had intended for the “robots.txt” specifications to form a part of the terms of service.[60]

 

V.             Prohibitions Of Contractual Overrides to Copyright Exceptions

 

The above analysis suggests that, even though the use of web scrapers could be permitted by copyright exceptions under certain circumstances, contractual restrictions might ultimately impose the same (or even more stringent) restrictions upon the use of web scrapers. For example, a user of a web scraper might be able to rely on the TDM exception to escape liability for copyright infringement, but a website’s terms of service might contain a restriction against web scraping, ultimately preventing the user from scraping the website. In this situation, the contractual restrictions imposed by the website owner have effectively overridden the copyright exceptions created by statute, rendering the copyright exceptions illusory (henceforth, these types of contractual restrictions shall be referred to as “contractual overrides”).

In response to these contractual overrides, some jurisdictions have legislated for statutory prohibitions which render contractual overrides generally unenforceable (henceforth, these types of statutory provisions shall be referred to as the “prohibitions of contractual overrides”). Where these prohibitions of contractual overrides apply to the TDM exceptions, as well as the “fair use” and “fair dealing” exceptions, they shall be discussed below.

 

A.    Contractual overrides to the text and data mining exceptions

 

It bears re-emphasis that the goal of the TDM exceptions is to remove the legal inhibitions created by copyright law in respect of TDM activities.[61] In that regard, allowing contractual overrides of the exceptions effectively leaves those who engage in TDM activities at square one.[62]

The UK and Singapore have responded by introducing blanket prohibitions of contractual overrides to their TDM exceptions.[63] In these two jurisdictions, the prohibitions of contractual overrides operate straightforwardly: they essentially provide that, to the extent that any contractual term is inconsistent with their respective TDM exceptions, the contractual term is rendered unenforceable.

Additionally, the EU provides an illustrative example of how a nuanced balance might be struck based on the relevant policy considerations, namely the promotion of innovation and ensuring freedom of contract: Under the EU CDSM Directive, the general TDM exception under Article 4 can be contractually overridden, but the specific exception under Article 3, which applies to TDM activities “conducted by research organisations and cultural heritage organisations” or “for the purposes of scientific research”, cannot be.[64] This represents the policy that TDM activities can “in particular, benefit the research community and, in doing so, support innovation”,[65] while for other organisations and purposes, “[r]ightsholders should remain able to license the uses of their works”.[66]

 

B.     Contractual overrides to the “fair use” and “fair dealing” exceptions

 

While a small number of jurisdictions feature blanket prohibitions in their copyright statutes against contractual overrides to most (or all) of their copyright exceptions,[67] specific prohibitions of contractual overrides to the “fair use” and “fair dealing” exceptions are rare.[68]

Jurisdictions such as Ireland adopt straightforward blanket prohibitions which apply to their “fair use” and “fair dealing” exceptions.[69] The prohibition of contractual overrides found in Singapore’s recently amended Copyright Act is also notable, as it provides for a robust set of requirements for contractual overrides to any copyright exception to be enforceable.[70] Of particular interest to users of web scrapers is the requirement that the contract must be individually negotiated before any contractual override may be enforceable; this requirement is unlikely to be made out for most contracts between the average users of web scrapers and website owners.

 

C.    Contractual restrictions on the use of non-copyrighted code?

 

At this juncture, a technical loophole remains under the copyright statutes which contain prohibitions to contractual overrides: some contractual restrictions might apply to both copyrighted and non-copyrighted content hosted on a website,[71] while the prohibitions of contractual overrides, on their face, only apply to copyrighted content.

For users of web scrapers, this potentially means that contractual liability could result from the web scraping of HTML code that does not satisfy the requisite threshold for originality.[72] In contrast, the web scraping of sufficiently-original HTML code would not result in any liability in copyright or contract. This state of affairs is plainly unsatisfactory, and this article suggests one workaround to it through the TDM exceptions.

It bears recalling that web scraping is a common precursor to TDM activities. Under such circumstances, the policy of the TDM exceptions, that TDM activities are beneficial to societies and economies,[73] applies equally to the mining of both copyrighted and non-copyrighted content. In this regard, the EU had considered clarifying that its TDM exception and corresponding prohibition of contractual overrides would be available for TDM activities concerning non-copyrighted content as well.[74] Therefore, it might be possible for the TDM exceptions to be interpreted by the courts to cover TDM activities concerning non-copyrighted content, though admittedly it would be awkward to interpret provisions in copyright statutes to cover uses of non-copyrighted material.

It would be much harder to extend the same argument beyond the TDM exceptions. For example, the “fair use” and “fair dealing” exceptions are ultimately premised on the general policy of copyright law,[75] and it would be conceptually unruly to extend prohibitions of contractual overrides of these exceptions over the web scraping of non-copyrighted content. In these situations, it might be possible to interpret the TDM exceptions even more broadly to encompass them, as the exceptions are non-exhaustively defined to begin with.[76]

 

VI.           Conclusion: What This Means For Users of Web Scrapers

Having considered the two copyright exceptions which are most relevant to web scraping, the potential contractual restrictions which could bind users of web scrapers, and how the prohibitions of contractual overrides to the copyright exceptions operate—what does it all mean for users of web scrapers?

As a starting point, the answer depends on which jurisdiction’s legal framework is at play. And it must be re-emphasised that the two legal controls discussed in this article—copyrights and contracts with websites—are not the only ones which exist in the web scraper’s world.[77] Nonetheless, in jurisdictions which provide users of web scrapers with both copyright exceptions and prohibitions of contractual overrides, it is submitted that, once the requirements for the relevant exceptions are met, web scrapers can be used without too much anxiety over liability for copyright infringement or breach of contract.

Of course, regardless of which legal frameworks are in place, web scrapers should always be used ethically. This exhortation is particularly important in the present age, where anyone with a computer and an Internet connection has easy access to the tools and tutorials needed to build their own web scrapers. Technological restrictions must be respected, “robots.txt” files should be followed, and when in doubt, it is always best to seek consent.[78] Through this article, it is hoped that those who wish to use web scrapers may do so with greater confidence in the legal frameworks underlying their activities—but without ever compromising our basic human decency.

 



* LL. B. (Hons) Candidate at the National University of Singapore; LL. M. Candidate at the University of Melbourne. This article would not have been possible without a proper appreciation of the technical details involved in web scraping—in that regard, I extend my most sincere gratitude to the Python programming communities on YouTube and StackOverflow for creating such wonderful programming tutorials, and making them easily and freely available. I also wish to extend the most heartfelt gratitude to Jerome Tan for the wonderful editing work and helpful comments on this piece. All errors remain my own.

[1] It is notable that “web scrapers” have been referred to in a variety of ways in the case law, such as “scraping programs”, “screen scrap[ers]”, “webcrawlers”, or “spider[s]”—see Andrew Sellars, "Twenty Years of Web Scraping and the Computer Fraud and Abuse Act" (2018) 24:2 BU J Sci & Tech L 372 at 381-382.

[2] Jay M Patel, Getting Structured Data from the Internet: Running Web Crawlers/Scrapers on a Big Data Production Scale, (California: Aspress, 2020) at xvii. For the avoidance of doubt, this article shall refer to web scraping algorithms (the bots themselves) as “web scrapers”, the ones who program or code those algorithms as “programmers of web scrapers”, and the human users of web scrapers (such as an app developer who uses a web scraper’s output to create an application) as “users of web scrapers”.

[3] A technical explanation of this process is provided in Part II of this article. Note that the term “webpage” has been preferred over “website”; from a technical perspective, a “webpage” refers to a single block of code that may be rendered on a web browser, while a “website” usually refers to a collection of webpages which are connected in some way—see “What is the difference between webpage, website, web server, and search engine?” (8 October 2021), MDN Web Docs (blog), online: <https://developer.mozilla.org/en-US/docs/Learn/Common_questions/Pages_sites_servers_and_search_engines>.

[4] Sellars, supra note 1 at 374.

[5] For an excellent technical summary of some real-world applications of web scrapers, see Patel, supra note 2 at 1-30.

[6] Christopher Watkins, “Web Scraping Fraud: Going, Going … Ongoing” (10 September 2019), Datavisor (blog), online: <https://medium.com/datavisor/web-scraping-fraud-going-going-ongoing-c0f7a0db7310>.

[7] Danny Palmer, “What is a DDoS attack? Everything you need to know about Distributed Denial-of-Service attacks and how to protect against them” (15 October 2020), ZDnet (blog), online: <https://www.zdnet.com/article/what-is-a-ddos-attack-everything-you-need-to-know-about-ddos-attacks-and-how-to-protect-against-them/>.

[8] See e.g. Singapore: Computer Misuse Act (Cap 50A, 2007 Rev Ed Sing), s 7; UK: Computer Misuse Act 1990 (UK), c 18, s 3; US: 18 USC § 1030 (2018).

[9] See Patel, supra note 2 at 60.

[10] In addition to statutes which deal with DDoS attacks, one major area of contention concerning the use of web scrapers is in personal data protection—see e.g. “Facebook says hackers ‘scraped’ data of 533 million users in 2019 leak”, The Straits Times (7 April 2021), online: <https://www.straitstimes.com/world/united-states/facebook-says-data-on-530-million-users-scraped-before-september-2019>. While it is an incredibly fascinating area of law to look into, it is, most unfortunately, beyond the scope of this article.

[11] The technical explanations in Part II of this article are based on the conventions used by Python programmers, as it is generally accepted that the Python programming language is the industry standard language for data science and machine learning—see e.g. “Why is the Data Science Industry Demanding Python?” (10 February 2020), Institute of Data (blog), online: <https://www.institutedata.com/blog/why-is-the-data-science-industry-demanding-python/>. However, similar processes should apply for other programming languages.

[12] Patel, supra note 2 at 31.

[13] The reader may experience this by opening any webpage and finding the “View page source” setting for the browser in use—this setting shows the underlying HTML code of the webpage that is currently open. It should be noted that this article only focuses on the legal protections surrounding the HTML code; the words, images and videos that result from the rendering of HTML code might be protected by different copyrights or contractual terms.

[14] Note that the technical details regarding accessing webpages through web browsers are much more complicated than this brief description might suggest—webpages are located at a specified “Internet Protocol Address”, which might be rendered in text form through the “Domain Name System”. These are standard protocols that the Web is run upon.

[15] Patel, supra note 2 at 37.

[16] Ibid at 41-42.

[17] Google has expended significant efforts in making the REP into an Internet standard—see Henner Zeller, Lizzi Harvey & Gary, “Formalizing the Robots Exclusion Protocol Specification” (1 July 2019), Google Search Central Blog (blog), online: <https://developers.google.com/search/blog/2019/07/rep-id>.

[18] On the difference between “webpages” and “websites”, see n 3 of this article.

[19] This is the author’s simplification; for a slightly-more technical description, see Patel, supra note 2 at 59.

[20] “Can I block just bad robots?”, The Web Robots Pages (blog), online: <https://www.robotstxt.org/faq/blockjustbad.html>.

[21] Patel, supra note 2 at 387.

[22] Google: “Introduction to robots.txt”, Google Search Central Documentation, online: <https://developers.google.com/search/docs/advanced/robots/intro>; Bing: “robots.txt tester”, Microsoft Bing Webmaster Tools help & how-to, online: <https://www.bing.com/webmasters/help/robots-txt-tester-623520ca>.

[23] Zeller, Harvey & Gary, supra note 17.

[24] “Advanced: How Search Works” (22 November 2021), Google Search Central, online: <https://developers.google.com/search/docs/advanced/guidelines/how-search-works>.

[25] See e.g. Daniel Seng, “The State of the Discordant Union: An Empirical Analysis of DMCA Takedown Notices” (2014) 18 Va JL & Tech 369.

[26] See Marrakesh Agreement Establishing the World Trade Organization, 15 April 1994, 1869 UNTS 154 Annex 1C at art 9(2) at art 10(1) (entered into force 1 January 1995) [TRIPS Agreement].

[27] See e.g. the facts of The Newspaper Licensing Agency Ltd v Meltwater Holding BV [2011] EWCA Civ 890 [Meltwater].

[28] See TRIPS Agreement, supra note 26 at art 9(2). For a domestic example, see Global Yellow Pages Ltd v Promedia Directories Pte Ltd [2017] SGCA 28 at para 15 [Global Yellow Pages].

[29] See e.g. Global Yellow Pages, supra note 28 at para 24; Feist Publications, Inc. v Rural Tel. Service Co., 49 US 340 at 348 [Feist]; ECJ Infopaq International A/S v Danske Dagblades Forening, C-5/08, [2009] ECR I-6624 at para 37.

[30] Feist, supra note 29 at 345; Global Yellow Pages, supra note 28 at para 27;

[31] Feist, supra note 29 at 347. Incidentally, this line of reasoning was suggested by respondents to the public consultations for the Singapore Copyright Act—see Ministry of Law Singapore & Intellectual Property Office of Singapore, Singapore Copyright Review Report (17 January 2019) at para 2.8.1.

[32] Computer Associates Intern., Inc. v Altai, Inc. 982 F (2d) 693 at 709-710 (2nd Cir 1992).

[33] EC, Policy Department for Citizens’ Rights and Constitutional Affairs, The Exception for Text and Data Mining (TDM) in the Proposed Directive on Copyright in the Digital Single Market – Legal Aspects (February 2018) at 12-13 [TDM – Legal Aspects].

[34] UK: Copyright, Designs and Patents Act 1988 (UK), c 48, s 29A [UK CDPA]; EU: Directives EC, Directive (EU) 2019/790 of the European Parliament and of the Council of 17 April 2019 on copyright and related rights in the Digital Single Market and amending Directives 96/9/EC and 2001/29/EC, [2019] OJ, L 130/92 at art 3 [EU CDSM Directive]; Singapore: Copyright Act 2021 (No 22 of 2021, Sing), s 244 [Singapore Copyright Act].

[35] See Ministry of Law Singapore & Intellectual Property Office of Singapore, Public Consultation on Proposed Changes to Singapore’s Copyright Regime (23 August 2016) at para 3.64.

[36] UK: Ian Hargreaves, Digital Opportunity: A Review of Intellectual Property and Growth (May 2011) at para 5.26; EU: TDM – Legal Aspects, supra note 33at 14-15; Singapore: Singapore Copyright Review Report, supra note 31 at para 2.8.4.

[37] UK CDPA, supra note 34, s 29A.

[38] Singapore Copyright Act, supra note 34, ss 243-244. This exception also applies to preparatory work done for the purposes of computational data analysis.

[39] A notable exception is Japan. See Tatsuhiro Ueno, “The Flexible Copyright Exception for ‘Non-Enjoyment’ Purposes – Recent Amendment in Japan and Its Implication” (2021) 70:2 GRUR International 145 at 149.

[40] UK: Intellectual Property Office, Exceptions to copyright: Research (October 2014) at 8; Singapore: Singapore Copyright Review Report, supra note 31 at para 2.8.5.

[41] See Sellars, supra note 1 at 373-375.

[42] Jonathan Band & Jonathan Gerafi, “The Fair Use/Fair Dealing Handbook” (March 2015), infojustice.org, online: <http://infojustice.org/wp-content/uploads/2015/03/fair-use-handbook-march-2015.pdf> at 3.

[43] There is considerable academic literature surrounding the proposition that the “fair dealing” and “fair use” exceptions are not as different as the language of their provisions might suggest—see e.g. Ariel Katz, “Debunking the Fair Use vs. Fair Dealing Myth: Have We Had Fair Use All Along?” in Shyamkrishna Balganesh, Ng-Loy Wee Loon & Sun Haochen, eds, The Cambridge Handbook of Copyright Limitations and Exceptions (Cambridge: Cambridge University Press, 2021) 111. However, a discussion on this issue would extend far beyond the scope of this article—it suffices to note that the issue exists.

[44] See e.g. UK: UK CDPA, supra note 34, ss 29, 30 and 30A; New Zealand: Copyright Act 1994 (NZ), 1994/143, ss 42-43.

[45] See e.g. Singapore: Singapore Copyright Act, supra note 34, s 191; US: 17 USC § 107.

[46] See e.g. Australia: IceTV Pty Ltd v Nine Network Australia Pty Ltd [2009] HCA 14 at para 24-25; Canada: Cinar Corp. v Robinson 2013 SCC 73 at para 23, [2013] 3 SCR 1168; US: Altai, supra note 32 at 711.

[47] See e.g. Singapore: Global Yellow Pages, supra note 28 at paras 74-76; US: Campbell v Acuff-Rose Music, Inc., 114 S Ct 1164 at paras 3,4 (1994).

[48] This is true even in jurisdictions which feature the narrower “fair dealing” exceptions in their copyright statutes—see e.g. UK: Ashdown v Telegraph Group Ltd [2001] EWCA Civ 1142 at para 70; New Zealand: Media Works NZ Ltd v Sky Television Network Ltd (2007) 74 IPR 205 at para 74.

[49] See e.g. UK: Meltwater, supra note 27; US: Field v Google Inc., 412 F Supp (2d) 1106 (D. Nev. 2006).

[50] See e.g. the clause at issue in HiQ Labs, Inc. v Linkedin Corp, 938 F (3d) 985 at 991, n 5 (9th Cir 2019) [Linkedin]: more than 95 million automated attempts to scrape data were blocked every day, based on the clause in the terms of service which provided that users were not allowed to “[s]crape or copy profiles and information of others through any means (including crawlers, browser plugins and add-ons, and any other technology or manual work)”, “[u]se manual or automated software, devices, scripts robots, other means or processes to access, ‘scrape,’ ‘crawl’ or ‘spider’ the Services or any related data or information”, or “[u]se bots or other automated methods to access the Services”.

[51] See e.g. the clause at issue in PropertyGuru Pte Ltd v 99 Pte Ltd [2018] SGHC 52 at para 77 [PropertyGuru]: the terms of service state that the site’s content can “only be used for your own and non-commercial use, and not for publication, distribution, transmission, retransmission, redistribution, broadcast, reproduction or circulation to someone else in the same company or organisation, and not for posting to other websites or forums, newsgroups, mailing lists, electronic bulletin boards, or Internet Relay Chats operated by other websites”.

[52] See Linkedin, supra note 50.

[53] This argument was attempted unsuccessfully in PropertyGuru, supra note 51. As a causal link between the alleged act of inducement and alleged breaches could not be established, the court did not see the need to examine the other elements of the tort. However, in the author’s view, the tort of inducement of breach of contract would be an appropriate cause of action in a situation involving an application built upon web-scraped data, where the web scraping constituted a breach of a website’s terms of service.

[54] See Meltwater, supra note 27 at para 49.

[55] Eliza Mik, “Contracts Governing the Use of Websites” (2016) Sing JLS 70 at 74-75.

[56] Kevin Conroy & John Shope, “Legal Analysis: Look Before You Click: The Enforceability of Website and Smartphone App Terms and Conditions” (2019) 63 Boston Bar J 23 at 23. Note also the categorical “click-wrap”, “browse-wrap” or “sign-in wrap” approach from US case law, which Professor Mik criticises for failing to examine the intention of the contracting parties—see Mik, ibid at 73-74.

[57] See “Can a /robots.txt be used in a court of law?”, The Web Robots Pages (blog), online: <http://www.robotstxt.org/faq/legal.html>.

[58] It is assumed that that the “robots.txt” file cannot constitute a contract in itself, either due to lack of consideration or lack of intention to create legal relations.

[59] See Thornton v Shoe Lane Parking Ltd [1971] 1 All ER 686 (a party is bound by terms which they either know of, or reasonable steps have been taken to give them notice of the terms; see also the famous “red hand rule”, on the same page) at 690.

[60] Carlill v Carbolic Smoke Ball Company [1983] 1 QB 256 at 261-262.

[61] See Part II.B of this paper.

[62] Singapore Copyright Review Report, supra note 31 at para 2.14.8. Note that this may be limited to particular situations where contractual overrides are likely to be prevalent—see e.g. TDM – Legal Aspects, supra note 33 at 13 (which notes that “[r]esearch and database providers often contractually override exceptions and limitations”).

[63] UK CDPA, supra note 34, s 29A(5); Singapore Copyright Act, supra note 34, s 187(1)(c).

[64] EU CDSM Directive, supra note 34, articles 3 and 4 read with article 7(1).

[65] Ibid, recital 8.

[66] Ibid, recital 18.

[67] See the examples cited in “Protecting Exceptions Against Contract Override: A Review of Provisions for Libraries” (27 November 2019), IFLA, online: <https://www.ifla.org/publications/node/92678> at 3.

[68] The author could not find any specific prohibitions of contractual overrides of the “fair use” and “fair dealing” provisions.

[69] Copyright and Related Rights Act, 2000 (No. 28 of 2000, Ireland), s 2(10).

[70] Singapore Copyright Act, supra note 34, s 186(2)(a).

[71] See e.g. the clauses extracted at n 50 and n 51 of this article.

[72] It bears re-emphasis that the issue of whether users of websites are contractually bound by these terms is controversial—see generally Part IV of this paper. At the very least, this is a source of legal uncertainty for users of web scrapers.

[73] See Part III.B.1 of this article.

[74] TDM – Legal Aspects, supra note 33 at 6.

[75] See Part III.C of this article.

[76] See Part III.B.1 of this article.

[77] See Part I of this article.

[78] When seeking consent from website owners, they might offer access to their application programming interfaces (“APIs”), which could operate more efficiently than web scrapers!

Empirical Legal Research in Singapore: Its Uses, Its Limitations, and the Way Forward

A PDF version of the article can be found here.


Empirical Legal Research in Singapore: ITS USES, ITS LIMITATIONS, AND THE WAY FORWARD

 

Darren Ang*

 

I.               Introduction

 

A.    Lawyers against scientists

 

Let us begin with a simple proposition: the methods used to reach conclusions in law are inherently different from those used in the hard sciences.[1]

We can explain this with a quick thought experiment. Ask a lawyer to prove a point, and they would probably point to an authoritative source that is broadly related to the conclusion sought to be reached, then hammer in their conclusion with arguments from principle and logic (and sometimes, rhetoric).

In contrast, ask a scientist to prove a point, and their approach is quite different. They would probably make a hypothesis, then run experiments to collect data before analysing that data using statistical methods. Their conclusions are reached through careful observation, complete with declarations of percentage-accurate degrees of confidence in their results.

If these contrasting methodologies had to be summarised in one word each, it could be said that methods in law are “argumentative”, while those in the hard sciences are “empirical”. More broadly, it could also be said that law is “qualitative" while the hard sciences are “quantitative”.


 

B.     The landscape of empirical legal research

 

However, over the past half-century, empirical methodologies have increasingly found their way into legal scholarship, particularly in the United States.[2] Entire textbooks have been written on the subject,[3] and legal studies with empirical components have come to take on bolder inquiries—including assessing judges’ behaviour and explaining their individual writing styles.[4]

In Singapore, the empirical charge has just begun to take hold. Over the past decade, at least six studies have relied on some form of empirical methodology, and all of them have relied on published judicial decisions as quantitative data.[5]

It turns out that there is a well-developed discipline in communications research which neatly encapsulates methodologies which “proceed from text to results”—its name is “content analysis”.[6] However, among the six empirical legal studies in Singapore, only the study by Lo et al has expressly acknowledged that it was adopting a content analysis methodology.[7]

In light of the recent empirical legal research charge, this article seeks to shed some light on the nature of content analysis methodologies, its limitations in the context of analysing published judicial decisions, and some potential workarounds to these limitations. It concludes with a brief suggestion that, in such contexts, the “empirical” methodologies of content analysis achieve their objectives best when paired with conventional, “argumentative” legal analysis.

 

 

II.             What is Content Analysis, and why is it appropriate For Legal Research?

 

A.    Content analysis and its methodology

 

Content analysis is a research technique that seeks to make replicable and valid inferences from texts to the contexts of their use.[8] That is, it seeks to draw meaningful conclusions through analysing large bodies of text quantitatively, such that future researchers applying the same methodology would reach the same conclusions.

The methodology of content analysis is robust—in a leading text on the discipline, Krippendorff identified six “components” of content analysis, which are as follows:[9]

1)     Unitizing: distinguishing segments of text which are of interest to an analysis;

2)     Sampling: limiting observations to a manageable subset of units that is statistically or conceptually representative of the set of all possible units;

3)     Recording/Coding: interpreting the unitized data and stating one’s experiences either in the formal terms of an analysis (recording) or according to observer-independent rules (coding);

4)     Reducing: using established statistical techniques or other methods for summarising or simplifying data;

5)     Abductively inferring contextual phenomena: bridging the gap between texts and what the texts imply using analytical constructs;

6)     Narrating: making the results comprehensible to others.

While a full exposition of each of these six components is out of the scope of this article, some discussion of the component of “abductive inference” is apposite. This component is said to “distinguish content analysis from other modes of inquiry”,[10] and it materialises as “analytical constructs” which function as “the best hypothesis or explanation that the analyst can imagine or defend … backed by knowledge of the context of the analysed texts”.[11]

The distinctive element of “abductive inference” makes content analysis methodologies particularly appropriate for the analysis of published judicial decisions—in this context, the “analytical constructs” can take the form of legal principles derived from conventional legal analysis, and these can be employed to justify quantitative findings.[12] To illustrate this with an example from an upcoming empirical legal study involving the author, a quantitative finding that a traffic offender’s plea of guilt is given mitigatory weight about 80% of the time may be explained with an argument from the sentencing objective of specific deterrence.

 

B.     The history and development of content analysis in legal scholarship

 

The earliest examples of content analysis are found in the quantitative analyses of printed matter by the Church in the 17th Century, to which the Church concluded that the printing of non-religious materials were a threat to its authority—this went on to inspire a significant 20th Century movement where various researchers engaged in quantitative analyses of newspapers and propaganda publications in an attempt to uncover, among other things, the profit motives behind newspapers and their negative effects on society.[13] Around that time, the seeds to the content analysis movement in legal scholarship were sown.

It has been said that “[t]he epistemological roots of content analysis [in legal scholarship] lie in Legal Realism”.[14] Legal Realism was a movement within the American legal academic circle that first gained traction in the 1920s,[15] and while the Realists departed from each other at various points, they shared a common scepticism towards conventional legal theories and zeal for reform.[16] Among them, a significant faction of the Realists sought to predict judges’ decisions with some degree of certainty,[17] and the empirical analysis of recorded judicial opinions was a particularly appropriate means towards that end.[18]

While the full extent of Realist thought has since lost most of its force,[19] the Realists’ clarion call to empiricism survived:[20] following the explosion of quantitative studies involving the content analysis of published judicial decisions in the United States in the 1990s-2000s,[21] the systematic content analysis of published judicial decisions is now “a mainstay of legal and political science scholarship”.[22]

 

C.    The place of content analysis in modern legal scholarship

 

What, then, is the place of content analysis in modern legal scholarship? It is said that content analysis “trades the pretence of ontological certainty for a more provisional understanding of case law”.[23] That is, conventional legal analysis requires the subjective, “deeply reflective” interpretation of a narrower area of the law,[24] while content analysis reaches an objective, “thinner” understanding of a large number of decisions.[25] They are different tools within the toolbox of legal analysis.

It follows that the role of content analysis in legal scholarship is not to supersede conventional legal analysis; instead, its role is to complement and augment conventional analysis.[26] For example, while conventional legal analysis is best suited for landmark judgments with great legal and cultural significance,[27] content analysis is particularly useful at “proving a negative”—if Principle Y states that Factor X will not be given weight except in exceptional cases, it can only be tested by looking through a sample of cases where Factor X was brought to the court’s attention, and finding that Factor X was given no weight in almost all of the sampled cases. It follows that both tools can be used in tandem to reach more robust conclusions[28]—continuing off the above example, if the content analysis reveals that Factor X is actually given weight in a significant proportion of cases, this would strongly support an argument for Principle Y to be reformed (or repealed).

However, there exists a more compelling reason to use content analysis in tandem with conventional legal analysis—there are several limitations inherent in published judicial decisions which, in most cases, render content analysis methodologies incapable of reaching robust conclusions in and of themselves.

 

III.           Limitations and potential workarounds

 

A.    The problem of unpublished decisions

 

Not every dispute goes to court, and the ones that do are often resolved without written or published opinions. In Singapore, the existence of unpublished decisions is well-known[29]—judges generally do not owe a duty to issue written grounds of decision. The most common situation for the duty to issue a written grounds of decision to arise, in both civil and criminal cases, is when a notice of appeal is filed.[30]

From these circumstances alone, it could be assumed that published decisions would likely involve more contentious cases, while straightforward cases are more likely to be unpublished.[31] However, this forms an insurmountable hurdle for aspiring researchers at the “sampling” component of content analysis, as the sample of published judicial decisions would never be representative of the whole population of interest of an empirical legal study. [32]

One workaround suggested by Hall & Wright is to acknowledge this hurdle and explicitly limit the scope of the study to published judicial decisions.[33] For example, in the study on the development of Singapore law by Goh & Tan, the authors limited their sample to reported cases, justifying this by arguing that reported cases “perhaps provide more significant influence on our local jurisprudence”.[34] In the author’s view, Goh & Tan’s argument sufficiently addresses the problem of unreported cases while also providing a positive justification for their sample.

However, even in studies where limiting the sample to published or reported cases cannot be similarly justified, it is said that a “skewed view” is better than having no view on the matter, and published decisions, as one of the significant sources of law for lawyers in the common law tradition, are still a “highly valuable source for systematic study”.[35] The limitations to empirical legal studies only mean that researchers must be “less expansive … in drawing conclusions from their findings”.[36] For example, while empirical studies measuring the effects of extra-legal factors on appellate decision-making in the United States had been subject to harsh attacks on their accuracy, more nuanced empirical legal studies that have acknowledged the limitations of their methodologies and employed more sophisticated techniques have been taken to more kindly.[37]

 

B.     The problem with analysing causative relationships between facts and decisions

 

In addition to the problem of unpublished decisions, Hall & Wright argue that a “circularity problem” arises when content analysis is employed to find causative relationships between legally relevant factors and judicial opinions, as the written facts and opinions may not fully capture the “real world facts” or the entirety of the case process.[38] This raises an issue at the “abductive inference” component of content analysis—abductive inference contemplates finding the best explanation to a particular set of facts, but if the facts themselves are incomplete, any inferences made from them will be similarly imperfect.

Unfortunately, to the author’s knowledge, no workaround is available to deal with this problem, and various empirical legal studies have faced harsh attacks on their validity for failing to take it into account.[39] While some empirical legal researchers have resorted to gathering data by physically attending court hearings,[40] and those studies have a stronger claim to their validity, no empirical study could possibly account for the closed-door and confidential nature of judicial decision-making.[41] Therefore, a similar attitude as with that towards unpublished decisions must be adopted: the problem must be acknowledged, and the conclusions sought to be reached must be restricted accordingly.

To that end, it is suggested that any empirical legal study that seeks to find the “weight” or “significance” attached to factors considered in judicial decisions may overreach the boundaries of content analysis—as “weight” or “significance” is a qualitative inquiry, best suited for conventional legal analysis. Most of the empirical legal studies in Singapore appear to have recognised this, and they have generally involved the counting of factors without any evaluation of causative significance, coupled with qualitative analyses of the findings using more conventional techniques of legal analysis.[42] In the author’s view, this combination the best balance between depth and objectivity of understanding.[43]

 

IV.           Conclusion

 

This article has set out the basic methodology and historical development of content analysis as a methodology in empirical legal scholarship, as well as its uses, limitations, and potential workarounds to those limitations. In summary, it has been argued that in the context of analysing published judicial decisions, content analysis methodologies achieve their objectives best when paired with conventional legal analysis, and while the problems with unpublished decisions and analysing causative relationships can be mitigated to some degree by such a pairing, they should be explicitly dealt with (or simply acknowledged) where they arise.

The empirical legal research charge has already begun—it should be welcomed with open arms. It is hoped that aspiring empirical legal researchers remain cognisant of the unique capabilities and limitations of their methodologies, and that the introduction of empirical methodologies into legal scholarship in Singapore will be the catalyst for fruitful discussions and developments in the law.



* LL.B. (Hons.) Candidate, National University of Singapore. The author wishes to express his most heartfelt gratitude to Mr. Benny Tan (Sheridan Fellow, National University of Singapore) for leading the empirical legal research charge within the Singapore Law Review, and for the many illuminating discussions about empirical methodologies in law—from which the seeds to a forthcoming empirical legal research paper and this companion piece were sown.

[1] For more variants on this illustration and the inspiration for this section, see Robert Lawless et al, Empirical Methods in Law, 2nd ed (Alphen aan den Rjin: Wolters Kluwer, 2016) at 7-20.

[2] Mark A Hall & Ronald F Wright, “Systematic Content Analysis of Judicial Opinions” (2008) 96:1 Cal L Rev 63 at 72.

[3] See eg, Lawless et al, supra note 1.

[4] James C Phillips & Edward L Carter, “Oral Argument in the Early Roberts Court: A Qualitative and Quantitative Analysis of Individual Justice Behaviour” (2010) 11:2 J App Pr & Pro 325; Keith Carlson et al, “A Quantitative Analysis of Writing Style on the U.S. Supreme Court” (2016) 93:6 Wash ULO 1461.

[5] These are: Goh Yihan & Paul Tan, “An Empirical Study on the Development of Singapore Law” (2011) 23 SAcLJ 176; Lee Zhe Xu et al, “The Use of Academic Scholarship in Singapore Supreme Court Judgments” (2015) 33 Sing L Rev 25; Cheah W L & Goh Yihan, “An Empirical Study on the Singapore Court of Appeal’s Citation of Academic Works: Reflections on the Relationship Between Singapore’s Judiciary and Academia” (2017) 29 SAcLJ 75; Agnes Lo et al, “The Evaluation of Medical Expert Opinions in Litigation: An Empirical Study” (2018-2019) 36 Sing L Rev 247; Jerrold Soh, “A Network Analysis of the Singapore Court of Appeal’s Citations of Precedent” (2019) 31 SAcLJ 246; and Professor Gary Chan’s monograph at Gary Chan Kok Yew, Tort of defamation before the Singapore Courts, 1965-2015: A comparative and empirical study (Singapore: Academy Publishing, 2017).

[6] Klaus Krippendorff, Content analysis: an introduction to its methodology, 2nd ed (California: Sage Publications, Inc., 2004) at 83

[7] Agnes Lo et al, “The Evaluation of Medical Expert Opinions in Litigation: An Empirical Study” (2018-2019) 36 Sing L Rev 247 at 254, 255.

[8] Krippendorff, supra note 6 at 18.

[9] Ibid at 83-85, 126, 171. Also note that these components do not have to be organised linearly, and a content analysis design can contain iterative loops: see ibid at 85.

[10] Ibid.

[11] Ibid at 171.

[12] See ibid at 90 for a research design framework that seeks to “operationalise expert knowledge”; this directly supports the use of legal principles as analytical constructs.

[13] Ibid at 3-6.

[14] Hall & Wright, supra note 2 at 76.

[15] Michael Freeman, Lloyd’s Introduction to Jurisprudence, 9th ed (London: Sweet & Maxwell, 2014) at 845.

[16] Ibid.

[17] Brian Z Tamanaha, Law as a Means to an End: Threat to the Rule of Law (Cambridge: Cambridge University Press, 2006) at 70.

[18] Karl Llewellyn famously read thousands of cases randomly selected from various American appellate courts to determine the factors which would influence a judge’s decision. See Karl N Llewellyn, The Common Law Tradition: Deciding Appeals (Boston: Little, Brown & Co, 1960).

[19] Tamanaha, supra note 17 at 1, 72.

[20] Hall & Wright, supra note 2 at 76.

[21] See Table 1 in Ibid at 72.

[22] Carlson et al, supra note 4 at 1466.

[23] Hall & Wright, supra note 2 at 87.

[24] Ibid.

[25] Ibid at 78.

[26] Ibid at 88. See also the concept of “triangulation” in the social sciences, in ibid at 83.

[27] Ibid at 84, citing the well-known case of Roe v Wade 410 U.S. 113 (1973).

[28] Ibid at 81.

[29] For a readily observable example, it appears that the Singapore courts have added the word “unreported” in brackets to indicate that a case cited as authority is an unpublished decision.

[30] For the position in the civil law, see Rules of Court (Cap 322, s 80, 2014 Rev Ed Sing), O 42 r 8(1). For the position in the criminal law, see Criminal Procedure Code (Cap 68, 2012 Rev Ed Sing) at ss 377(5) and (7). See also ss 394A-B and 397(3A) of the Criminal Procedure Code for more exceptional situations where a written grounds of decision must be issued.

[31] This assumption is backed by some literature. Judge Edwards wrote that “any assessment of the work of the courts of appeals that does not include unpublished decisions cannot be seen as complete”, and that according to official statistics, less than 17 percent of all opinions in courts of appeals were published. See Harry T Edwards & Michael A Livermore, “Pitfalls of Empirical Studies that Attempt to Understand the Factors Affecting Appellate Decisionmaking” (2009) 58:8 Duke U 1895 at 1923. In Singapore, this is likely to be the case as well—as an illustration, the State Courts had heard 303,487 criminal cases in the year 2018 alone, while searching the term “Public Prosecutor” in LawNet yields 15,482 results across all years. See “One Judiciary Annual Report 2018”, Supreme Court Singapore, online: <https://www.supremecourt.gov.sg/docs/default-source/default-document-library/ojar_full-8.pdf>.

[32] In Singapore, there is at least one channel for researchers to access court archives—the Empirical Judicial Research Programme. However, there appear to be no channels for application for this programme that are available to the general public. See “About the Empirical Judicial research Programme”, Singapore Judicial College, online: <https://www.supremecourt.gov.sg/sjc/empirical-judicial-research>.

[33] Hall & Wright, supra note 2 at 92.

[34] Goh Yihan & Paul Tan, “An Empirical Study on the Development of Singapore Law” (2011) 23 SAcLJ 176 at para 43.

[35] Hall & Wright, supra note 2 at 92.

[36] Harry T Edwards & Michael A Livermore, “Pitfalls of Empirical Studies that Attempt to Understand the Factors Affecting Appellate Decisionmaking” (2009) 58:8 Duke U 1895 at 1907.

[37] Ibid at 1904-1905.

[38] Hall & Wright, supra note 2 at 95.

[39] See Edwards & Livermore, supra note 36 at 1930-1944.

[40] See eg, Jessica Jacobson & Mike Hough, “Personal Mitigation: An Empirical Analysis in England and Wales” in Julian V Roberts eds, Mitigation and Aggravation at Sentencing (Cambridge: Cambridge University Press, 2011) at 146-167.

[41] Edwards & Livermore, supra note 36 at 1903.

[42] A notable exception is Lo et al, supra note 8, which the author is immensely grateful to have been involved in—the methodology of the study involved various coders identifying certain factors and assigning “scores” from 1 to 5 to them based on their “significance”. While a critical examination of the methodology used is out of the scope of this article, it suffices to say that this is a novel methodology which warrants much closer examination. See also Jerrold Soh, “A Network Analysis of the Singapore Court of Appeal’s Citations of Precedent” (2019) 31 SAcLJ 246, the first empirical legal study to adopt network analysis in Singapore.

[43] Hall & Wright, supra note 2 at 88.

Torn Fealty to the Courts and Science: Conundrums over Medical Expert Opinions (II/II)

The PDF version of this article is available for download here.


Torn Fealty to the Courts and Science: Conundrums over Medical Expert Opinions (II/II)

Agnes Lo, Bryont Chin, Darren Ang, Leon Tay & Louis Lai*

The previous section of this article examined the different considerations that affect the weight accorded to expert evidence.1 It was observed that judges scrutinize the expert himself (his qualifications and the level and relevance of his specialization) and also his testimony (its cogency, internal consistency, and consistency with the facts).

A survey of local case law and overseas publications reveals issues surrounding this assessment. These have been directed to the reliability of expert opinions itself, applying generally to all experts. Unfortunately, local jurisprudence has not addressed these concerns.

The lacuna in local jurisprudence on this point is especially unfortunate given the widespread prevalence of and reliance upon medical expert opinion in Singapore. It is thus pertinent to examine these issues and their causes in detail.

I. ISSUES ARISING FROM THE ADVERSARIAL PROCESS

A. Partiality and bias of experts

Case law reflects a persistent concern about the partiality and bias of experts but also a recognition that this will persist as long as litigation remains adversarial. Ideally, under Order 40A Rule 2(2) of the Rules of Court2 [ROC], experts should bear an exclusive duty to the Court; however, this is usually not the case in practice. A “market-place” mentality continues to prevail: parties select experts because the latter’s views are “already known and, consequently, would advance the party’s case.”3 Both the courts and commentators like Professor Pinsler have acknowledged that the source of this paradigm is the adversarial model, since parties assume that their “remuneration of the expert justifies a measure of loyalty that will somehow manifest to his advantage in the determination of the case before or at trial”.4 Therefore, as long as the adversarial model continues to be adopted, parties will continue to pay, retain, and offer the experts future engagements—all of which incentivise experts to provide opinion evidence slanted in favour of the clients who engage them.5

To make matters worse, breaches of the procedural duty under Order 40A above attract no direct penalties in themselves. There is only an evidential penalty: if an expert is found to be partial, the court completely or partially disregards his testimony.6 This does not seem to have been enough: from the introduction of Order 40A in 2000 to date, the courts have continued to observe bias in experts.7 Therefore, bias among experts remains a real problem in the court today.

B. A fear of testifying in Court

Also flowing from the adversarial model and the process of adjudication itself is medical experts’ intrinsic fear of testifying in court. Experts—and medical experts in particular—tend to steer clear of the witness stand for two reasons: a fear that they may not be sufficiently competent to testify, and a fear that they may be shamed or wronged on the stand.8

While there is no equally comprehensive study, local doctors have expressed the same fears of testifying due to natural embarrassment and a fear of damaging their colleagues’ reputations.9 Failing to address these fears will restrict plaintiffs’ access to medical experts or otherwise make engaging experts more expensive. This reluctance will affect the quality and accessibility of court adjudication, reliant as it is on expert testimony.

C. Insufficiency of factual bases for medical opinions

Experts often have insufficient facts to support their opinion. This is because in litigation, they are likely to be engaged before trials begin since their opinions are critical in establishing the client’s case and thus predict the likelihood of success at trial.10 However, this means that much of the evidence the expert relies upon to develop his report and testimony will be “untested raw material” since this evidence has yet to be tested in court.11 This concern came to a head in Khoo Bee Kiong,12 where the court expressly questioned the factual basis of an expert opinion prepared with affidavits of evidence in chief.

This is likely to be aggravated in medical negligence claims, where the recent Supreme Court Practice Directions13 impose an accelerated case management timeline. As an expert report must now be filed before the claim itself begins, medical experts have less time and evidence to formulate their opinions.

II. ISSUES IMMANENT IN THE JUDICIAL TREATMENT OF MEDICAL EXPERT OPINIONS

Apart from issues with the expert opinions themselves, there are also notable concerns with the judicial treatment of medical expert opinions as reflected in empirical studies. These suggest that there are systemic issues with judges’ ability to assess the reliability and credibility of a medical expert’s opinion and the proper weight to be accorded to admitted evidence.

A. Lack of clarity on the assessment of reliability and credibility of medical expert opinions

At its core, the court’s assessment of the reliability and credibility of a medical expert’s opinion is a discretionary exercise. It is naturally opaque and applied circumstantially, and may hence appear uncertain. This means that the other stakeholders involved—lawyers, clients, and the medical experts themselves—may not fully understand the standard expected of them. Jurs, for instance, observes that legal professionals tend to over-estimate the value of personality and credentials, whereas judges place a greater emphasis on partiality and bias, tentativeness, and the technicality of the opinions proffered.14 Clarity in this area is much needed;15 consensus and better articulation of the relevant considerations would assist greatly.

B. Assessment of weight to be accorded to admitted medical expert opinions

Beyond a lack of clarity on a systemic approach to assess reliability and credibility of medical expert opinions, there is a concern that the current approach may mean that factors which might appear to be unrelated or insignificant may be taken into account in the assessment of weight to be accorded to admitted expert opinions. For instance, “impressions of the analyst’s demeanour and credibility, like the ability to survive cross-examination, will not in most cases provide rational means of assessing the probative value of an opinion.”16 The English courts, for example, have been criticised for assessing expert opinion evidence on this basis, which has led to factually unsustainable acquittals.17

Studies suggest that this concern may not be unfounded because of the way medically-untrained judges and jurors assess the value of medical expert opinions. For instance, jurors focus on consensus supporting the expert opinion, whether the opinion is applicable in the case at hand,18 and how experts communicate their opinions.19 However, these indicators may not necessarily go towards the veracity of an opinion.20 Where earlier studies have suggested that this may be due to a lack of education in the expert’s field,21 it is understandable why some doctors feel that laypersons like jurors “should not decide medical malpractice cases because of the arcane issues involved in the practice of medicine.”22

Local jurisprudence has only engaged these concerns in the abstract. Ronald Wong posits that the law’s insistence on the finality of an opinion for it to be accepted as evidence conflicts with the scientific method, which relies not on finality, but rather social acceptance to establish a theory as valid.23 This mirrors the observation in Levett and Korvera’s review that “individuals often are unable to evaluate statistics or methodology properly… [such that] it is reasonable to assume that jurors may be unsuccessful in independently detecting flaws in research presented by an expert in court.”24 Nonetheless, Wong has argued that it is still preferable for a judge to continue to adjudicate on the issue, and if the expert evidence is insufficiently reliable, a judge ought to make his findings of fact on the burden of proof.25 Wong’s argument—defended on grounds of a public interest in resolving disputes—contrasts with Professor Hor’s view that judges lack any institutional capacity to determine the veracity of expert opinion.26

Since judges must still decide cases involving expert opinions, the question of institutional competence remains a live issue to be managed. This is especially the case if adjudication may lead to imperfect reliance on wrong or inaccurate expert opinions as suggested by the studies above.

III. THE PRESENT STATE OF MEDICAL EXPERT OPINION AND ITS PRACTICAL USE

Concerns over unreliable expert opinions have been the subject of statutory reform. The government has astutely institutionalised a court-appointed panel of psychiatrists testifying in criminal cases under the new s 270 of the Criminal Procedure Code27 [CPC] which has not yet come into force.28 This places the control over the admission and use of psychiatrists’ opinions in the hands of a Selection Committee instituted for these purposes.

In other aspects, however, reform continues to be lacking. The independence of experts continues to be secured by weak duties under s 269 CPC which do not depart from the phraseology of Order 40A ROC. The amendments to the CPC do not address the deeper concerns relating to adversarial litigation or pronounce on the feasibility of alternatives such as independent experts or assessors.

Therefore, the fundamental question remains as to whether the amendments will indeed improve the reliability of psychiatric evidence and be a viable option across other classes of medical experts as well; or, alternatively, whether it would entrench a preferred doctrine of psychiatric testimony.29 If the latter is the case, then the amendments may exacerbate the issue of expert witness bias. Psychiatrists have questioned why they were the only class of expert that required supervision, while the Criminal Bar expressed that it would be preferable if the selection process also included defence counsel as a member of the Selection Committee.30

Beyond the concerns regarding the partiality of experts, there are also practical complications of note. Amendments have been made to the Supreme Court Practice Directions to require medical expert opinion reports to be submitted in the pre-writ exchange of information for medical negligence cases, but this may mean that experts will not have sufficient factual evidence to support their opinions.31 These remain with broader concerns that judges may lack the institutional capacity to accurately assess medical expert opinions, or that they may demand a different standard from non-medical expert witnesses. While there are extra-legal methods employed to cope with these such as a training course for medical experts to raise the quality of medical opinions and reports, the impact of these courses on the judiciary’s assessment of the medical opinions and reports presented to them, has not yet been observed.

IV. CONCLUSION

The discussion above has traced a number of difficulties in the use of medical experts’ opinions, as well as the use of experts in general. The judicial approach to assessing the probative value and reliability of opinions is plagued with difficulties stemming from intractable concerns that medical experts tend to be partial towards the party that hired them, and that such experts often lack sufficient facts to properly develop an opinion. Although there are both existing procedural laws and external support to safeguard the reliability of medical expert opinions, these are often watered down by parties’ adversarial mentalities. Where there is little incentive to cooperate, parties do not actively rely on the same measures such as concurrent expert evidence or discussion. Other issues, such as fear among medical professionals of proffering testimony against their fellow doctors, or the institutional competence of the Court to adjudicate on the same, have hardly been the subject of discussion by the legal community in Singapore.

This is unfortunate as expert opinions are the lifeblood of many disputes. Even before a case commences, expert opinions are important in managing client expectations, where the expert’s assessment indicates the client’s likelihood of success.32 Medical experts are of special significance in the field given that they are relevant in many legal disputes and, in fact, comprise the bulk of experts sought.

Meeting the concerns above would require lawyers, judges, and medical experts alike to come to a clear consensus on their roles and competencies in adjudication. Coping with the difficulties of an adversarial model will also require a change in mindset and culture: lawyers and courts must emphasise to medical experts that in testifying, they are helping society as a whole. In this regard, the authors posit that empirical studies will be helpful in assessing the gravity of each issue, and the effectiveness of measures to cope with them. In this regard, this article will be followed by a further submission, “The Evaluation of Medical Expert Opinions in Litigation: An Empirical Study”,33 which seeks to provide more clarity on the factors that judges in Singapore apply in evaluating medical expert opinion evidence, and their relative significance in different scenarios.


*The authors are a group of undergraduate students in the National University of Singapore, and are presently in the course of an empirical research project on the subject.

[1] See Agnes Lo, Bryont Chin, Leon Tay & Louis Lai, “Torn Fealty to the Courts and Science: Conundrums over Medical Expert Opinions (I/II)”, online: Juris Illuminae, Vol 10 <http://www.singaporelawreview.com/juris-illuminae-entries/2018/torn-fealty-to-the-courts-and-science-conundrums-over-medical-expert-opinions-i>.

[2] Cap 322, R 5, 2014 Rev Ed Sing.

[3] Jeffrey Pinsler, “Expert Evidence and Adversarial Compromise: A reconsideration of the Expert’s Role and Proposals for Reform” [2015] 27 Sing Ac LJ 55 at [2].

[4] Ibid at [15].

[5] Vita Health Laboratories Pte Ltd v Pang Seng Meng [2004] 4 SLR 162, at [81].

[6] If the expert attempts or is seen to be an advocate for his party’s cause, he will inexorably lose credibility: per VK Rajah JA in Vita Health Laboratories Pte Ltd v Pang Seng Meng [2004] 4 SLR 162 at [83].

[7] Supra note 3 at [4].

[8] See e.g. “Medical Malpractice - Expert Testimony” (1965-1966) 60 Nw UL Rev 834.

[9] See Joseph Sheares, “Writing the Expert Report and Testifying in Court (Part 2)”, SMA News (February 2015) at 23, online: <https://www.sma.org.sg/UploadedImg/files/Publications%20 %20SMA%20News/4702/Professionalism.pdf>.

[10] Amanda Stevens, “Editorial: Reliability and cogency of expert witness evidence in modern civil litigation” (2011) 66 Anaesthesia 764.

[11] Khoo Bee Keong v Ang Chun Hong [2005] SGHC 128 at [72].

[12] Ibid.

[13] See Sing, The Supreme Court Practice Directions (2017) part XXIII, s 158(1), referring to Appendix J (High Court Protocol for Medical Negligence Cases), online: <https://epd.supremecourt.gov.sg/downloads/Appendix_J/APPENDIX_J.pdf> at 4.1-4.2.

[14] Andrew W Jurs, “Expert Prevalence, Persuasion, and Price: What Trial Participants Really Think About Experts” (2016) 91 Ind LJ 353 at 371-372, 377-378 and 389.

[15] See e.g. the Ministry’s comments in proposing amendments to s 270 of the CPC, which it expressed to be for the purpose of ensuring that evidence given by psychiatrists is “competently arrived at and objective”: Siau Ming En, “Proposed psychiatric panel must be large enough for smoother defence: Lawyers”, Today (28 July 2017), online: <https://www.todayonline.com/singapore/proposed-psychiatric-panel-must-be-large-enough-smoother-defence-lawyers>.

[16] Gary Etmond, “Legal versus non-legal approaches to forensic science evidence” (2016) 20:1 IJEP 3 at 24.

[17] Ibid.

[18] Lora M Levett & Margaret Bull Korvera, “The Effectiveness of Opposing Expert Witnesses for Educating Jurors about Unreliable Expert Evidence” (2008) 32:4 Law and Human Behaviour 363-374, at 364.

[19] Andrew W Jurs, “Expert Prevalence, Persuasion, and Price: What Trial Participants Really Think About Experts” (2016) 91 Ind LJ 353.

[20] Kovera MB, McAuliff BD & Hebert KS, “Reasoning about Scientific Evidence: Effects of Juror Gender and Evidence Quality on Juror Decisions in a Hostile Work Environment case” (1999) 84:3 J Appl Psychol 362 at 365.

[21] See e.g. Kalven, Harry Jr & Hans Zeisel, The American Jury (Chicago: University of Chicago Press, 1966) at 153; cited in Sanja Kutnjak Ivkovic & Valerie P Hans, “Jurors’ Evaluations of Expert Testimony: Judging the Messenger and the Message” (Cornell Law Faculty Publications, Paper 385, 2003) at 443.

[22] Neil Vidmar, “Lay Decision-Makers in the Legal Process” in Peter Cane & Herbert M Kritzer, The Oxford Handbook of Empirical Legal Research (OUP, 2010) 626, at 633.

[23] Ronald JJ Wong, “Judging between Conflicting Expert Evidence” (2014) 26 Sing Ac LJ 169.

[24] Supra note 18.

[25] Supra note 23.

[26] Ibid, citing Michael Hor, “When Experts Disagree” (2000) Sing JLS 241 at 243.

[27] Cap 68, 2012 Rev Ed Sing.

[28] Clause 78 of the Criminal Justice Reform Bill (Bill 14 of 2018), which introduces s 270 of the CPC.

[29] Supra note 15.

[30] Ibid.

[31] See Appendix J of the Supreme Court Practice Directions (Amendment No. 3 of 2017). The relevant amendments to the Practice Directions are available online at <http://www.supremecourt.gov.sg/docs/default-source/default-document-library/pd-amd-no-3-of-2017.pdf?Status=Temp&sfvrsn=0.9324472121860925>.

[32] Supra note 10.

[33] Lo et al., “The Evaluation of Medical Expert Opinions in Litigation: An Empirical Study” (2018–2019) 36 Sing L. Rev. 247.


Footpath Warriors: A Proposed Sentencing Framework for Personal Mobility Device Accidents

The PDF version of this article is available for download here.


Footpath Warriors: A Proposed Sentencing Framework for Personal Mobility Device Accidents

Darren Ang*

I. INTRODUCTION

The use of personal mobility devices [PMDs] on footpaths in Singapore has led to numerous PMD-related accidents,1 some of which have resulted in grave injuries to the victims.2 The propensity for PMDs to cause harm is well known by the public – there have been numerous calls to tighten legislative controls on PMD usage, including a request to completely ban PMD usage on footpaths.3

These calls have not gone unanswered, and Parliament has been consistently tightening legislative controls over PMD usage over the past year. The passing of the Active Mobility Act 20174 [AMA] added s 5A to the Road Traffic Act5, which expressly prohibits the use of PMDs on public roads in most circumstances. Following the AMA, the Active Mobility Regulations 20186 introduced additional controls over, inter alia, the sale and modification of PMDs.

While these legislative controls are relatively new, and their effectiveness cannot be conclusively determined, it unfortunately appears that the measures in place are still insufficient – calls for the complete banning of PMDs have not ceased.7 To supplement the legislative measures already in place, the courts may step in to pass sentence on PMD users who cause hurt to pedestrians, and in doing so, deter such careless behaviour enough to lower the incidence of PMD accidents.

II. SCOPE AND PURPOSE OF THE ARTICLE

This article proposes a sentencing framework for PMD accident cases where grievous hurt is caused, by reconciling the unique characteristics of PMD accident cases with the sentencing framework for causing grievous hurt by a negligent act that endangers human life under s 338(b) of the Penal Code8 [PC] in Tang Ling Lee v Public Prosecutor9 [Tang Ling Lee]. The offence of causing grievous hurt by a negligent act that endangers human life is chosen as a benchmark, as most reported PMD accidents would likely fall under this provision.10

III. THE TANG LING LEE FRAMEWORK

In Tang Ling Lee, See Kee Oon J laid down a general sentencing framework to be applied for road traffic cases charged under s 338(b) of the PC when the accused claims trial (referred to in this article as the ‘Tang Ling Lee framework’).11 The Tang Ling Lee framework is a two-step inquiry, which first categorises the offence under one of three pre-established categories to determine a starting point sentence (referred to in this article as the ‘three-category approach’). Following which, the framework allows further adjustments to be made to take into account the relevant mitigating and aggravating factors (referred to in this article as the ‘adjustments step’).12

At the first step of the inquiry, a presumptive sentencing range will be determined as a starting-point sentence, having regard to the twin considerations of harm and culpability.13 See Kee Oon J summarised the three-category approach in the following table format:14

  • Category 1:

    • Circumstances: Lesser harm and lower culpability;

    • Presumptive Sentencing Range: Fines.

  • Category 2:

    • Circumstances: Greater harm and lower culpability Or Lesser harm and higher culpability;

    • Presumptive Sentencing Range: One to two weeks’ imprisonment.

  • Category 3:

    • Circumstances: Greater harm and higher culpability;

    • Presumptive Sentencing Range: More than two weeks’ imprisonment.

See Kee Oon J defined “harm” as the “nature and degree of the grievous bodily injury caused to the victim(s)”15, and the “degree of culpability” as “the degree of relative blameworthiness disclosed by an offender’s actions … measured chiefly in relation to the extent and manner of the offender’s involvement in the criminal act”.16 Additionally, the factors which would affect the “culpability” limb under the three-category approach would include: (i) the manner of driving, (ii) the circumstances which might have increased the danger to road users, and (iii) the offender’s reasons for driving.17

IV. THE TANG LING LEE FRAMEWORK ADJUSTED FOR PMD ACCIDENTS

While the Tang Ling Lee framework was caveated to only apply to road traffic cases,18 it has been adjusted to apply in non-road traffic cases as well. Most notably, the District Judge in Public Prosecutor v Cai Mei Ying19 [Cai Mei Ying] applied the three-category approach under the Tang Ling Lee framework for a s 338(b) case involving a bicycle accident, but held that the presumptive sentencing ranges were not binding due to the different contexts in which the cases occurred.20

It is submitted that when a PMD accident case charged under s 338(b) of the PC reaches the courts, there would be no practical reason to deviate from the approach taken by the District Judge in Cai Mei Ying. Moreover, the District Judge in Cai Mei Ying noted that the Tang Ling Lee framework was expressly caveated to only apply to road traffic cases, but recognised that the parties had agreed that “because Tang Ling Lee involves the co-existence of vehicles and humans in shared spaces”, the framework was applicable to their case.21 PMD accidents also involve the “co-existence of vehicles and humans in shared spaces”, and it would follow that the approach taken in a future case involving a PMD accident is likely to be similar to the approach taken by the District Judge in Cai Mei Ying.

It would then be apposite to consider what adjustments might be made to the Tang Ling Lee framework for PMD accident cases, and it is argued that the main considerations for adjusting the Tang Ling Lee framework in these cases would be: (i) the lower propensity for PMDs to cause harm when compared to motor vehicles (referred to in this article as the ‘harm factor’), and (ii) the objective of general deterrence (referred to in this article as the ‘deterrence factor’). If the courts were to follow the approach in Cai Mei Ying for adjusting the Tang Ling Lee framework, they would apply the three-category approach without using the presumptive sentencing ranges, and then account for the above two considerations at the adjustments step of the Tang Ling Lee framework. These two considerations will now be dealt with in turn.

A. Downward adjustments for the ‘harm factor’

PMDs have a lower propensity to cause harm than motor vehicles, as they travel at lower speeds and are not as heavy as motor vehicles. This could warrant a downward adjustment of the starting-point sentence at the adjustments step of the Tang Ling Lee framework.

As discussed above, the approach in Cai Mei Ying ought to be the first port-of-call for the discussion. However, the District Judge in Cai Mei Ying did not make a clear finding on the effect of the ‘harm factor’, though it was noted that as a guiding principle, “drivers of heavy vehicles stand to receive heavier punishments than riders of light vehicles due to the greater damage their vehicles can cause”.22

It is then helpful to consider the approach taken in Public Prosecutor v Khairul bin Hairuman23 [Khairul], which was a case involving a fatal bicycle accident. The accused in Khairul was charged under the rashness limb of s 304A of the PC, and the District Judge applied the three-category approach from the case of Public Prosecutor v Ganesan Sivasankar24 [Ganesan], including the presumptive sentencing ranges.

The Ganesan framework was also laid down by See Kee Oon J to apply to any s 304A rashness case when the accused claims trial, and it uses a three-category approach followed by an adjustments stage as with the Tang Ling Lee framework.25 The only differences between the two frameworks are that the Ganesan framework does not consider the harm caused (as it is, by definition of the offence, the death of the victim),26 and that the presumptive sentencing ranges in Ganesan under the three-category approach are more severe. The table laid out by See Kee Oon J in Ganesan is illustrative of these differences:27

  • Category 1:

    • Accused’s culpability: Low;

    • Presumptive sentencing range: 3 to 5 months’ imprisonment.

  • Category 2:

    • Accused’s culpability: Moderate;

    • Presumptive sentencing range: 6 to 12 months’ imprisonment.

  • Category 3:

    • Accused’s culpability: High;

    • Presumptive sentencing range: More than 12 months’ imprisonment.

The District Judge in Khairul found that the accused’s conduct fell within Category 2 of the Ganesan framework, and then discounted the accused’s sentence at the adjustments stage to “reflect the lower consciousness of risk of harm towards others associated with the riding of the bicycle in comparison with the riding or driving of motorised vehicles”.28

The approach in Khairul would support the position that the lower propensity for PMDs to cause harm could count for a significant downward adjustment at the adjustments stage of the Tang Ling Lee framework.

B. Upward adjustments for the ‘deterrence factor’

As discussed in the introduction of this article, PMD usage is an issue of great public concern in Singapore, and deterrent sentencing in the courts would be an appropriate supplement for the legislative controls already in place. Therefore, the court may be inclined to make an upward adjustment of the starting-point sentence for the purpose of general deterrence.

In Cai Mei Ying, the ‘deterrence factor’ was effectively the only aggravating factor taken into consideration at the adjustments stage of the Tang Ling Lee framework.29 Interestingly, the District Judge was cognisant of the issue of PMD accidents, classifying both cyclists and PMD users within the same class of persons and commenting that “with the increased popularity of PMDs and the use of bicycles … all cyclists and users of PMDs must be reminded to take extra care when they are in shared spaces”.30 Therefore, it is likely that the court would take the ‘deterrence factor’ as a significant aggravating factor at the adjustments stage of the Tang Ling Lee framework.

Additionally, a common thread that binds most cases charged under s 338(b) of the PC is that the accused person had breached certain safety regulations which led to the accident – the accused in Cai Mei Ying was cycling in a no-cycling zone31 and the accused in Tang Ling Lee failed to give way to a motorist with the right of way32. The breach of safety regulations counted towards a finding of higher culpability at the three-category approach stage, ultimately leading to a higher starting-point sentence. Therefore, it could be argued that the purpose of general deterrence is served indirectly at the “culpability” limb of the three-category approach.

V. CONCLUSION

To summarise the proposed method for adjusting the Tang Ling Lee framework for PMD accident cases charged under s 338(b) of the PC, the courts are likely to apply the three-category approach without using the presumptive sentencing ranges to determine a starting-point sentence. Following which, at the adjustments stage, the court would account for the ‘harm factor’ to adjust the starting-point sentence downwards and the ‘deterrence factor’ to adjust the starting-point sentence upwards.

For PMD users, who would bear the full brunt of this proposed method, the message is clear – if ever placed in an unfortunate s 338(b) situation where the harm has already been caused, it would be in their best interest to render as much assistance to the victim as necessary and comply with all orders, for a possible finding of lower culpability.


* A most heartfelt thanks to Professor Alan Tan, for answering my queries and helping immensely with the framing of my research questions.

[1] Adrian Lim, “Parliament: About three accidents a week involving personal mobility device users” (8 January 2018), The Straits Times, online: <www.straitstimes.com/politics/parliament-average-of-three-accidents-a-month-involving-pedestrians-and-personal-mobility>.

[2] Shaffiq Idris Alkhatib, “Teen e-scooter rider pleads guilty in incident which caused pedestrian severe brain injuries” (13 June 2018), The Straits Times, online: <www.straitstimes.com/singapore/courts-crime/teen-e-scooter-rider-pleads-guilty-in-incident-which-caused-pedestrian-severe>.

[3] Desmond Ng & Kan Lau, “Why being hit by an e-scooter can be deadly – and a call to ban them from footpaths” (20 May 2018), Channel NewsAsia, online: <www.channelnewsasia.com/news/cnainsider/e-scooter-ban-footpaths-accidents-safety-registration-debate-10250946>.

[4] No 3 of 2017, Sing.

[5] Cap 276, 2004 Rev Ed Sing.

[6] S 251/2018 Sing.

[7] Rodney Tan, “It Is Time to Ban E-Scooter[s] in Singapore” (last updated 11 December 2018), ipetitions, online: <www.ipetitions.com/petition/it-is-time-to-ban-e-scooter-in-singapore>.

[8] Cap 224, 2008 Rev Ed Sing.

[9] [2018] SGHC 18; [2018] 4 SLR 813.

[10] For example, see supra, notes 1-3.

[11] Supra note 9 at [32].

[12] Ibid.

[13] Ibid.

[14] Ibid at [31]. The information is represented in table format in the PDF version of this article.

[15] Ibid at [25].

[16] Ibid.

[17] Ibid at [27].

[18] Ibid at [24].

[19] [2018] SGMC 56.

[20] Ibid at [23].

[21] Ibid at [20].

[22] Ibid at [24].

[23] [2018] SGMC 16.

[24] [2017] SGHC 176; [2017] 5 SLR 681.

[25] Ibid at [54].

[26] Ibid.

[27] Ibid at [55]. The information is represented in table format in the PDF version of this article.

[28] Supra note 23 at [38].

[29] Supra note 19 at [37].

[30] Ibid.

[31] Ibid at [2].

[32] Supra note 9 at [5].