The United Kingdom’s overhaul of intermediary liability : the role of policy discourse in explaining sudden policy change

(1)

Master thesis Political Science

Public Policy & Governance

June 2018

The United Kingdom’s overhaul of intermediary liability:

the role of policy discourse in explaining sudden policy change

Submitted to:

Supervisor:

Dr. Rosa Sanchez Salgado

Assistant Professor of

European Public Policy at the

University of Amsterdam

Second reader:

Dr. Imrat Verhoeven

Assistant Professor of Public

Policy at the University of

Amsterdam

Submitted by:

Storm Gibbons

10789685

storm.gibbons7@gmail.com

University of Amsterdam –

Graduate School of Social

Sciences

Word count: 23500

Date: 22-06-2018

(2)

1

1. Introduction

Online platforms are unmistakable components of contemporary society. They connect users to their social and professional networks, enable transactions between buyers and sellers through efficient matching of supply and demand, facilitate ‘sharing’ of under-utilised assets and online ‘collaboration’, and provide access to a wealth of content and information. The potential for online platforms to monetise these kinds of intermediary activities, enabled by advancements in technology, has accelerated the rise of the ‘platform economy’. The platform economy is a relatively novel concept that describes all economic activities facilitated by online platforms, but that has only slowly gained traction over the past decade – although online platforms have been around since the turn of the millennium. The platform economy comes with high growth potential, as it captures “additional value for [its] participants, value that was previously out of reach” (Coyle 2016). However, as online platforms often play an intermediary role, assigning liability for illegal and harmful content shared by their users presents a complex challenge. While representatives of online platforms argue that their role is passive and that end-users should therefore solely be liable for such content, opponents maintain that online platforms should have adequate safeguards in place to prevent, detect, and take down inappropriate material. Previous research has examined intermediary liability, primarily from a legal perspective. Research has focused on the application of the landmark E-Commerce Directive (ECD), which the European Union (EU) adopted in 2000 and remains the most important piece of legislation governing intermediary liability in the EU. However, recent policy developments at both national and EU level reflect a shift in government approaches towards stricter regulation of platform liability, for which the jurisprudence of national and European courts does not adequately account. Existing literature on Internet regulation does not specifically deal with the regulation of platform liability either, instead concentrating on the established culture of self-regulation governing the Internet, with which these developments are in conflict. Furthermore, academic literature that purportedly examines government responses to the platform economy (de Groen et al. 2017; Lenaerts, Beblavý & Kilhoffer 2017; Dølvik & Jesnes 2017), actually only covers specific subsets of the platform economy, like the ‘sharing economy’, ‘collaborative economy’, or ‘gig economy’.

This thesis resolves this conceptual confusion and tackles the topic of intermediary liability from a political perspective, focusing on policy change, in order to explain the shift in government approaches to the intermediary liability of online platforms for illegal and harmful content. This is an unprecedented approach to this topic, which provides readers with a better understanding of policy developments regarding platform liability. The EU’s current

(4)

3 intermediary liability policy regime leaves much up to the discretion of Member States, resulting in divergent national approaches. In fact, the United Kingdom’s (UK) approach to intermediary liability has changed drastically since Theresa May became prime minister two years ago: from initiating a joint letter to the European Commission emphasising the opportunities of online platforms and warning against further regulation of platforms in April 2016 – just before May’s first ministry started – to committing to a review of its intermediary liability regime and possible further regulation of online platforms through the launch of its Digital Charter in January 2018. During my internship at Access Partnership, a technology policy consultancy firm, I learned that this change in approach had not been expected: not only because it was unprecedented, but also because “this process has been driven towards legislating without any real idea of what will go in a bill and [has taken] many in industry by complete surprise” (Dickson 2018). In order to explain this turnaround, this thesis conducts an in-depth case study focusing on the development of the UK’s government approach to online platforms and UK policy discourse on the topic of illegal and harmful online content, as this is at the centre of discussions of platform liability at the UK and EU level.

This thesis seeks to answer the question of whether prime minister Theresa May’s approach to platform liability constitutes policy change and if so, to what extent policy discourse on illegal and harmful online content explains this. To answer this question, it conducts a case study of policy discourse on illegal and harmful online content in the UK from May 2015 to May 2018. It examines policy discourse in order to understand policy change, as “speaking of change, (…) rather than just thinking it, is key to explaining the actions that lead to major policy transformations” (Schmidt 2011). This research contributes to a growing body of academic literature that conceives of discourse as an institution that affects policy change, referred to as discursive institutionalism. In doing so, this thesis analyses framing within policy discourse in order to determine frame alignment between different policy frames and a master frame. Processes of frame alignment explain support for varying levels of government intervention and this underpins government policy change (Barbieri 2015). The framing analysis focuses on contributions from key policy actors in the UK’s platform economy: HM government, major political parties, industry representatives, and non-governmental organisations (NGOs). It covers both ‘coordinative discourse’ between these actors – parliamentary debates, policy consultation contributions, and official letters – and ‘communicative discourse’ between policy actors and the public: laws, official statements, reports, publications of political parties (i.e. election material), and speeches. It contextualises these findings through brief consideration of alternative explanation for policy change, such as the actions of states, government change, and focusing events.

(5)

4

2. Literature Review

The platform economy is a novel area of academic study. Previous research has discussed what the platform economy comprises, highlighted challenges in regulating it, and examined legal issues arising from their business models. Research has focused on ‘intermediary liability’, which describes the (legal) relationship between online platforms and their third-party users. Understanding this relationship is crucial to understanding different national and transnational regulatory approaches to the platform economy. While legal studies into these concepts are plentiful, hardly any political research has been conducted into Member State approaches to online platforms and more specifically, platform liability for illegal and harmful content. This section emphasises the importance of further academic research into government policy approaches to platform liability. It resolves conceptual confusion about the ‘platform economy’, discusses challenges governments face in regulating online platforms, and reviews existing academic literature on intermediary liability.

2.1 A conceptual jungle: disentangling the meaning of the ‘platform economy’

Governments are faced with a constant struggle to keep up with technological developments, because any technical or conceptual confusion may lead to misinformed or misguided policies that either inhibit further innovation or do not adequately protect citizens against technology. Conceptual confusion is evident within many academic and professional accounts of the platform economy, as many authors mistakenly refer to the ‘platform economy’, ‘sharing economy’, ‘collaborative economy’, and even the ‘digital economy’ interchangeably. These kinds of conceptual errors have led to different conclusions regarding the size of the platform economy (Schmid-Drüner 2016) and absence of consensus on this issue, as demonstrated by the recent efforts of the Centre for European Policy Studies1_{. This section helps researchers navigate} this conceptual jungle by disentangling intertwined concepts related to the platform economy.

Online platforms

Both the conceptual errors and lack of agreement on the size of the platform economy are primarily down to the absence of an accepted definition of online platforms. Before delving into the concept of ‘online platforms’, it is important to note that the platform economy is a topic that falls within the EU’s Digital Single Market policy, where the EU and Member States have ‘shared competence’. As the EU was first to define online platforms and as EU law pre-empts

1_{The Centre for European Policy Studies organised an expert workshop in April 2018 to develop} common tools for researchers to measure its size and impact (CROS 2018). The outcomes of this workshop are yet to be published.

(6)

5 national law in areas of shared competence, academic research into the definition of online platforms largely focuses on the EU level. Specifically, the EU put forward a legal definition of ‘online platforms’: “an undertaking operating in two (or multi-) sided markets, which uses the Internet to enable interactions between two or more distinct but interdependent groups users so as to generate value for at least one of the groups” (European Commission 2015). Critics, however, considered the scope of the definition too broad and vague, and questioned the necessity of a definition of ‘online platforms’, given that no definition of ‘offline platforms’ exists (Gawer 2016). In particular, they argued that any intermediary operating online could be considered an online platform by that definition.

As many continued to apply the aforementioned rejected definition, a European Risk Observatory discussion paper on protecting workers in the online platform economy put forward a neutral, user-centric definition of online platforms: “online spaces where users can obtain information or interact socially or economically” (Garben 2017). While the definition remains relatively vague, it encompasses the “wide range of activities including online advertising platforms, marketplaces, search engines, social media and creative content outlets, application distribution platforms, communications services, payment systems, and platforms for the collaborative economy” (European Commission 2018a). The definition covers online intermediaries, but acknowledges that they are not necessarily platform-based. Furthermore, this definition also incorporates non-commercial interactions, by excluding any reference to markets, and omits online platforms’ level of passiveness in facilitating access to information and interactions between users. As determining the passiveness of online platforms is central to intermediary liability policy regimes across the EU, this thesis proceeds with the use of this definition for ‘online platforms’.

The platform economy

This thesis now compares the concepts of the “platform economy”, “sharing economy”, “collaborative economy”, and “digital economy” to create clarity within this conceptual patchwork. The platform economy broadly refers to “the entire economy comprising, involving and resulting from [online platforms]” (Garben 2017). The platform economy is considered a more neutral term than ‘sharing economy’, ‘collaborative economy’, and ‘gig economy’ (Forde et al. 2017; Kenney & Zysman 2016; Schmid-Drüner 2016). As the ‘collaborative economy’ and ‘sharing economy’ are considered to have overly positive connotations, while the ‘gig economy’ is negatively loaded, use of these concepts suggests support or opposition to these online platforms (Forde et al. 2017). As “there is a clear case for adopting a term as neutral as possible”

(7)

6 (Schmid-Drüner 2016), stakeholders increasingly prefer using the concept of ‘platform economy’ to other similar terms (Forde et al. 2017).

The platform economy also encompasses the ‘collaborative economy’, ‘sharing economy’, and ‘gig economy’. One application of online platforms in the platform economy is being “the method of (organising the) provision, sale and delivery of goods and services” (Garben 2017). The collaborative economy refers to “business models where activities are facilitated by collaborative platforms that create an open marketplace for the temporary usage of goods or services often provided by private individuals” (European Commission 2016b). As collaborative platforms organise the provision, sale, and delivery of temporary access to goods or services, the collaborative economy can be considered an application of the platform economy (Garben 2017). The key difference between the concepts of the ‘sharing economy’ and the ‘collaborative economy’ is that the former focuses on P2P interactions, while the latter comprises business-to-business (B2B) interactions, business-to-consumer (B2C) interactions, and peer-to-peer (P2P) interactions (Dervojeda et al. 2013). From this perspective, the ‘sharing economy’ can be considered a part of the ‘collaborative economy.

Similarly, the ‘collaborative economy’ can be considered a part of the ‘sharing economy’. The sharing economy involves “consumers (…) granting each other temporary access to their under-utilized physical assets ("idle capacity"), possibly for money” (Frenken et al. 2015; Meelen & Frenken 2015) and has four different manifestations: the access economy, the gig economy, the communing economy, and the collaborative economy (European Committee of the Regions 2015). Furthermore, the concepts of ‘sharing economy’ and ‘collaborative economy’ both incorporate the ‘gig economy’, which specifically emphasises the opportunities for independent workers to make use of these kinds of online platforms. Like the ‘collaborative economy’, the ‘sharing economy’ also relies on online platforms and is considered part of the broader platform economy. The EU’s consultation on online platforms confirms this, identifying both ‘collaborative economy’ and ‘sharing economy’ platforms as relevant online platforms (House of Lords 2015). Furthermore, the platform economy can be considered part of the wider digital economy: the economy based on ICT technologies (BCS 2013; L’Hoest 2001; Mondekar 2017). Summarising, both the ‘collaborative economy’ and the ‘sharing economy’ are part of the ‘platform economy’, which in turn is an essential constituent of the broader ‘digital economy’.

2.2 Regulatory challenges in the platform economy

Online platforms bring enormous value to the digital economy, driving innovation, facilitating (commercial) interactions, increasing consumer choice, improving industry competitiveness and

(8)

7 efficiency, and strengthening participation in society (European Commission 2018a). Furthermore, the platform economy stimulates entrepreneurship and non-traditional forms of work, including gig work (Forde et al. 2017; Schmidt 2017), as online intermediary platforms allow entrepreneurs to test business ideas on a small scale, and independent workers to find flexible work. Online platforms match supply and demand efficiently, help overcome market failures through peer-review trust mechanisms, and create indirect network effects, which serve both buyers and sellers in two-sided markets (Coyle 2016). Overall, the platform economy benefits businesses, independent workers, consumers, and general users.

Challenges within the platform economy include most dominant platforms being US-based and storing international users’ data in the US, the complexity of the platform architecture and underlying algorithms, and the power relationship between platforms and users (Helberger, Pierson & Poell 2018). Other disruptive effects include harm to traditional matchmaking industries, such as taxi companies and travel agents (Evans & Schmalensee 2016), and a lack of protections for workers (Fabo, Karanovic, & Dukova 2017), in the case of online labour platforms. The biggest challenge of all is the question of how to regulate online platforms. The regulation of the platform economy is a complex issue for policy-makers facing a “moving target” due to rapid technological developments (Munkøe, 2017). The rapid pace of change renders policy proposals irrelevant almost immediately (Claffy & Clark 2014). This leads to a high threshold for regulatory intervention, as regulators find themselves in a so-called Collingridge dilemma: intervening too early in technological developments might result in pre-emptive regulation that risks harming innovation, while intervening too late may mean that a technology can no longer be regulated effectively (Lynskey 2017). If regulators introduce broad horizontal regulation that is “not anchored in robust foundational principles, there is greater risk of manipulation of regulators and regulatory capture” (O’Connor & Schruers 2016). All in all, this means that introducing policies too hastily, without adequate scrutiny and input, will put them at risk of creating regulatory paralysis or regulatory capture (Lynskey 2017), but taking too much time will render them obsolete.

This is further complicated through the contrasting perspectives of different policy actors on the impact of online platforms and the effect ‘disruptive innovation’ on markets: “a process by which a product or service takes root initially in simple applications at the bottom of a market and then relentlessly moves up market, eventually displacing established competitors” (Clayton Christensen 2018). While the rise of the platform economy may hurt industry incumbents, academics have typically argued for limited regulation of online platforms – if any (Malhotra & van Alstyne 2014; Ranchordas 2015; Rauch & Schleicher 2015). Many authors are

(9)

8 entirely against government intervention altogether (Allen & Berg 2014; Cohen & Sundararajan 2015; Thierer et al. 2015). National governments in the EU seem to share these sentiments: they apply existing regulatory frameworks to the activities of online platforms (Lenaerts, Beblavý, & Kilhoffer 2017), created for comparable services offered by incumbents, and are formulating more sophisticated and elaborate policy responses to the platform economy (de Groen et al. 2017; Dølvik & Jesnes 2017; Lenaerts, Beblavý, & Kilhoffer 2017). However, the EU has identified that such national approaches may lead to an uneven playing field, as they often result in platforms being subject to less stringent rules than industry incumbents.

2.3 Intermediary liability for illegal and harmful online content

Online platforms can become hotbeds for illegal and harmful content, necessitating discussions on regulation. Platforms are vulnerable to illegal and harmful content due to their passive, intermediary role and can therefore easily be abused by criminals, terrorists, and ‘internet trolls’. Content may be harmful to minors, proliferate hate speech, incite terrorism, or infringe intellectual property rights (European Commission 2016a). Furthermore, platforms may be used to spread disinformation, or ‘fake news’. Previous research has mainly examined intermediary liability from a legal perspective and more recently the impact of EU policy reforms on its intermediary liability regime. Hardly any political research has been conducted into Member States’ intermediary liability regimes, with national approaches to illegal and harmful content on online platforms in particular remaining unchartered territory. In order to create the conditions for such political research, this section considers the definitions of illegal and harmful content, as well as synthesising and reviewing academic literature on intermediary liability.

Illegal and harmful content online

The EU has referred to the concepts of ‘illegal content’ and ‘harmful content’ in an online context since 1996. As it would be dangerous to conflate illegal and harmful content (European Commission 1996) this thesis now sets out both definitions. Illegal content refers to types of material posted online that infringes rules “which limit for different reasons the use and distribution of content” (European Commission 1996) and is generally legislated for at the Member State level. Examples of such content include child pornography, racist or xenophobic hate speech, incitement to terrorism, illegal commercial practices, and breaches of intellectual property rights (European Commission 2018b). Harmful content refers to types of (legal) material posted online that “may offend the values and feelings of other persons: content expressing political opinions, religious beliefs or views on racial matters etc.” (European

(10)

9 Commission 1996). Examples of harmful content include abusive material (i.e. bullying or trolling), extremist material, sexual or violent material, and ‘fake news’. As illegal and harmful content cover a wide range of inappropriate materials, academic research has tended to examine generic regulatory approaches to specific forms of content (i.e. hate speech), covering both online and offline manifestations, at the EU or national level.

Intermediary liability

Liability, in its most general sense, refers to legal obligations that natural or legal persons have towards each other. Intermediary liability therefore refers to the legal obligations intermediaries have towards their users. More specifically, it involves a breach of law, attributable to an intermediary service provider, that causes harm to a user of the intermediary service2_{. Many} different kinds of liability can be used to describe intermediary liability: primary liability, secondary liability, tertiary liability, third party liability, indirect liability, gatekeeper liability, or accessory liability (Angelopoulos 2016). Furthermore, as online platforms facilitate different kinds of illicit content and activities, from copyright infringements and fake news to abuse and incitement to terrorism, intermediary liability cuts across different kinds of law, such as public, private, property and criminal law.

However complex this may seem, the premise is simple: online platforms can be liable for the information they transmit or store, even when their role is merely intermediary. Whether or not a breach of law that takes place on an online platform is attributable to the platform depends on their level of passiveness. However, it is important to remember that online platforms do not necessarily constitute intermediaries nor do intermediary service providers necessarily constitute online platforms. The OECD proposes the following definition of online intermediary service providers: “Internet intermediaries bring together or facilitate transactions between third parties on the Internet. They give access to, host, transmit and index content, products and services originated by third parties on the Internet or provide Internet-based services to third parties” (OECD 2011). It is therefore clear that intermediary liability does not apply to all online platforms, nor is it exclusive to them.

The EU intermediary liability regime centres around the ECD. Previous research has examined this landmark legislation extensively, often focusing on ambiguities within those provisions of the ECD which provide certain kinds of intermediaries exemption from liability and how different Member States have implemented the Directive (Baistrocchi 2003; DLA Piper

2_{Adapted from the EU’s elements of state liability: “(i) a breach of EU law; (ii) attributable to the} Member State; (iii), which causes damage to an individual” (Eurofound 2011).

(11)

10 2009; iLINC 2015; Julià-Barceló & Koelman 2000; Svensøy 2011; Wendehorst 2016). Other authors have provided insights into the development of the EU regulatory framework and Court of Justice of the European Union (CJEU) case law regarding intermediary liability (Angelopoulos 2015; Angelopoulos 2016; Headdon 2012; Marusic 2016; van der Sloot 2015). Research also often concerns specific aspects of the EU’s intermediary liability regime, such as its notice-and-action initiative (Angelopoulos & Smet 2016; Kuczerawy 2015), its implications for different kinds of illegal content, like copyrighted material (Angelopoulos 2016; Hays 2006; Rosati 2016; Synodinou 2015), and content regulation (Akdeniz 2010). Furthermore, previous research has compared the EU and US intermediary liability regimes, as well as Member States’ regimes, from a legal perspective (Angelopoulos 2016; DLA Piper 2009; Julià-Barceló 2000). In other words, there is clearly a wealth of legal analysis of the EU’s and Member States’ intermediary liability regimes.

Contrarily, there is a lack of political research into intermediary liability regimes across the EU. The main difference between political and legal analyses of intermediary liability regimes is that political research carefully considers policies in light of broader trends in the field of internet regulation, such as “turning online intermediaries into internet policies” (Frosio 2017). Existing research focuses on recent or desired EU reforms (Frosio 2016; Frosio 2017; Stalla-Bourdillon 2017), the economic impact of its current policy approach (Ghafele 2016), and how ‘voluntary’ EU measures are shifting policy discourse from ‘intermediary liability’ to intermediary responsibility’ (van Eecke & Ooms 2007; Frosio 2018; Horten 2016a). The fact that political analysis into intermediary liability has initially focused on the EU level makes sense as the ECD, the bedrock of the EU’s intermediary liability regime, applies across Member States.

However, academic research has hardly paid any attention to Member States’ approaches to online platforms and in particular, intermediary liability for illegal and harmful content on online platforms. Furthermore, previous research has not yet determined whether policy discourse is shifting from intermediary liability to intermediary responsibility at the national level as well. Intermediary responsibility is a new concept which suggests that intermediaries should on their own initiative take action to protect their users (Horten 2016b). The ECD currently disincentivises platforms from taking action by granting platforms exemption from liability based on their levels of passiveness (Duckworth 2017). However, this approach “might be pushing an amorphous notion of responsibility that incentivizes intermediaries’ self-intervention to police allegedly infringing activities on the internet” (Frosio 2016). This thesis therefore closely examines the intermediary liability policy regime of the United Kingdom and national policy discourse on illegal and harmful content.

(12)

11

3. Theoretical framework

Building on the research gaps identified in the previous section, this thesis has formulated the following research question: Does prime minister Theresa May’s approach to platform liability constitute policy change and if so, to what extent does UK policy discourse on illegal and harmful online content explain this? In order to answer this question, the research question is broken down into three smaller subquestions, which find their foundations within this theoretical framework:

I. Determining policy change:

Can a policy change be determined in the UK’s approach to the intermediary liability of online platforms for illegal and harmful content since Theresa May became prime minister?

At first glance, it seems that the UK has drastically changed its approach to the regulation of platform liability drastically under Theresa May and that this happened relatively unexpectedly. This thesis therefore expects to determine that Theresa May brought about major policy change, which made the UK’s intermediary liability regime considerably stricter, very suddenly.

II. Frame alignment and policy change legitimation:

How does framing by key policy actors and specifically ‘frame alignment’ explain this policy change? As discourse on platform regulation and intermediary liability has primarily centred on the topic of illegal and harmful content online, this thesis examines discourse on this topic and specifically statements of key policy actors in order to determine their frame alignment with the established ‘master frame’. This thesis expects that this analysis of discourses will provide key insights into the interactions between the government and other key policy actors, helping to explain why the government decided to bring about such major and sudden policy change. While the policy change caught industry by surprise, it expects to identify frame alignment between key actors’ policy frames and the established ‘master frame’, which guided the government to change its regulatory approach to platform liability.

III. Other factors in the legitimation process:

How do alternative explanations of policy change account for the UK’s renewed regulatory approach to platform liability?

(13)

12 As Member States like France and Germany have also started to regulate online platforms more strictly, this thesis expects that Theresa May’s governments may have drawn inspiration from them in setting their new policy direction. It also expects that the government change after the 2017 general election provided further legitimation for policy change: despite the loss of its majority in Parliament, remaining the biggest party provided Theresa May with the mandate to deliver on her manifesto promises. In addition, key political events within the timeframe of the case study, such as terrorist attacks in the UK and the Cambridge-Analytica data scandal, may have changed the political mood in the country.

This chapter provides theoretical background to these subquestions. Section 3.1 explains what constitutes policy change and the link between different levels of policy change and types of regulation. Section 3.2 discusses different theoretical approaches to determining and explaining policy change, highlighting that these approaches fail to adequately explain how interactions between key policy actors affect policy change taking place during a short time period. Section 3.3 puts forward discursive institutionalism as an approach that succeeds in capturing this, as discourses of different policy actors may pressure decision-makers to take a different regulatory approach and bring about policy change. Section 3.4 discusses how framing analysis can be used to operationalise this discursive institutionalist approach and how frame alignment feeds into this. Finally, section 3.5 explains how this theoretical framework and the subquestions will be applied throughout the analysis.

3.1 Policy change

Policy change can refer both to “merely incremental refinements of earlier policies” and policies that are entirely new or innovative (Bennett & Howlett 1992). Policy change can be distinguished from policy reform in that policy reform generally refers to major or fundamental shifts in policy or regulation (Cerna 2013), whereas policy change covers a far wider spectrum of adjustments. The study of policy change therefore provides insights into the motivations for and effects of following different policy pathways. Alongside understanding whether a clear need for policy change exists, it is also important to understand to what extent policies can change. Sabatier distinguished three levels of policy change, namely: deep core, policy core, or secondary belief systems (1999). Deep core beliefs refer to normative assumptions about human nature and society, such as the role of government and how fundamental values are prioritised, which are very difficult to change (Sabatier 2006). Policy core beliefs are “applications of deep core beliefs

(14)

13 to an entire policy subsystem” (Sabatier 2006), which may be adjusted based on the needs of that specific subsystem. In practical terms, this means that politicians may find government intervention undesirable, yet necessary within certain policy subsystems: i.e. the platform economy, internet regulation, or intermediary liability. Secondary beliefs cover policy choices with a smaller scope and require less evidence to change (Sabatier 2006).

This thesis seeks to determine and explain changes in the UK government’s regulatory approach to platform liability, which involves explaining varying levels of government intervention regarding this policy topic at different times in the case study. Its analysis therefore focuses on changes to policy core of the UK’s intermediary liability policy regime and uses government intervention as a guide to identify these changes. Government intervention can be determined based on different forms of regulation: self-regulation, co-regulation, and state-regulation. As the Conservative Party has been in power for the last eight years and its fundamental view on government intervention has not changed, the analysis does not cover changes to the deep core. While identifying changes to deep core beliefs or policy core beliefs can be straightforward, changes to secondary aspects can be harder to determine. Furthermore, changes to secondary aspects of policies in most cases will not significantly affect the relationship between key policy actors. It therefore only discusses changes to secondary aspects of the government’s liability regime, where easily observable and appropriate.

Sabatier’s typology is a useful tool for understanding the dimensions of policy change, but it is important to note that there is no such typology covering the intensity of policy change. As change is always measured in relation to preceding periods of continuity, the intensity of policy change ultimately depends on context and can best be determined through normative evaluation. Nevertheless, it can be useful to compare the intensity of different policy changes in a consistent way, so this thesis distinguishes three scales of intensity: minor, moderate, and major policy change. To demonstrate how this relates to Sabatier’s levels of policy change, you could have minor policy change to deep core beliefs – i.e. from opposing government intervention in markets entirely to only supporting government intervention in the case of a financial crisis – but also major policy change to secondary beliefs: for example, ceasing to hold public consultations on internet regulation.

This thesis highlights UK policy measures that affect the government’s level of intervention in the platform economy to determine policy change to the core of the UK’s intermediary liability policy regime and discusses the intensity of this policy change. Furthermore, it analyses the EU’s intermediary liability regime to better understand the UK’s room for manoeuvre regarding its regulation of platform liability: how much scope did the UK

(15)

14 have to bring about policy change? Understanding the development of both the UK and EU intermediary liability regimes therefore provides context to the question of whether UK policy discourse explains the government’s policy change.

3.2 Determining and explaining policy change

This section discusses different theoretical approaches that help determine and explain policy change. Popular explanations of policy change include the incrementalist approach, the punctuated equilibrium (PE) model, the multiple streams framework (MSF), and the advocacy coalition framework (ACF). The incrementalist approach determines policy change when small, incremental changes to a policy over a long period of time lead to a broader policy change and/or that the policy no longer serves its intended purpose (Lindblom 1965; Cruz-Rubio 2011). As such, this approach determines policy change through closely examining the implementation of a policy in relation to its original design. The PE model complements the incrementalist approach, as it seeks to explain policy change that is not incremental but rather marks a large-scale departure from the past (True, Jones & Baumgartner 2006). Within this context, large-large-scale departures refer to the suddenness of policy change rather than the extent to which a policy has changed. The PE model implies that long periods of policy continuity (stasis) are followed by rapid spurs of dramatic policy change (punctuations). Both the incrementalist approach and PE model are therefore more useful for determining than explaining policy change.

Conversely, the MSF and ACF are more useful for explaining policy change. On the one hand, the multiple streams framework (MSF) conceives of policy as a product of three independent processes: a problem stream, a policy stream, and a politics stream. When these streams align, there is a ‘window of opportunity’ (Kingdon 1984). Windows of opportunity describe whether ‘political opportunity structures’ – institutional and political conditions that restrict possibilities for policy change (Meyer & Staggenborg 1996: 1634) – are receptive to policy change and if so, policy entrepreneurs can take advantage of this to put forward their policy solutions. This approach draws on the ‘garbage can’ model, which explains that decision-making, and therefore policy change, take place through ‘organised anarchy’, where problems and solutions are formulated independently and are dumped in a shared, figurative garbage can (Cohen, March & Olsen 1972). Taking together MSF and the garbage can model, this suggests that policy change is sometimes brought about by policy entrepreneurs “wait[ing] for the right time to present ready-made solutions” (Cairney 2013). On the other hand, the ACF focuses specifically on the interactions of different stakeholders in bringing about policy change. It thus builds on the theory of network governance, which explains that non-governmental stakeholders

(16)

15 and informal interactions play an increasingly influential role in policy-making and decision-making processes. Specifically, the ACF involves coalitions of actors with shared beliefs competing with each other to dominate policy-making in subsystems and produce policy change (Jenkins-Smith & Sabatier 1994). The MSF and ACF thus provide important insights into opportunities for and support for policy change.

However, these frameworks do not provide a comprehensive explanation of policy change within this case study for two main reasons. Firstly, the ACF deals with policy change over several decades, while the topic of platform liability for illegal and harmful online content has only really taken off in the past five years in the UK. Secondly, the MSF explains receptiveness to policy change but does not account for competition and interactions between different policy entrepreneurs. Other factors that partly, but not entirely, explain policy change include “the actions of other states or the discourse surrounding this issue” (Walsh 2006), as well as elections and corresponding changes in government. In addition, political mood for policy change can change through ‘focusing events’ – “sudden shocks to policy systems that lead to attention and potential policy change” (Birkland & Warnement 2014) – but this too only offers a partial explanation, as these events merely set policy change on the agenda. Once agenda-setting has taken place, it is up to policy actors to gain support for the solutions they put forward so that policy change is ultimately brought about. This thesis therefore only briefly discusses the explanations mentioned above. Instead, it analyses policy discourses to explain the government’s policy change, as speaking of change explains actions leading to major policy change (Schmidt 2011) and this provides insights into the interactions between key policy actors on specific policy topics and above all, policy change.

3.3 Discursive institutionalism

This thesis focuses on UK policy discourse to explain the government’s policy change regarding platform liability. As most policy developments in this field so far concern illegal and harmful online content, it specifically focuses on UK policy discourse on illegal and harmful online content. In order to avoid any confusion, it is important to stress the difference between discourse analysis and analysis of discourses – this thesis subscribes to the latter approach – although it must be noted that this distinction should not be drawn too sharply (Bacchi 2009). While discourse analysis examines how actors use language and specifically draw upon ‘discursive structures/meanings’ to work their way through different situations (Stapleton & Wilson 2004), analysis of discourses refers the study of meaning production within specific social settings, based on institutionally supported and culturally influenced interpretations and concepts

(17)

16 (Bacchi 2009). As such, analysis of discourses is characterised by “the project of identifying and analysing ‘discourses’ within texts” (Bacchi 2009: 22). Concretely, this may involve researchers conducting textual analysis of policy materials, such as manifestos or speeches, in order to identify and explain patterns within and between documents.

Policy discourse comes in two forms: “the coordinative discourse among policy actors and the communicative discourse between political actors and the public” (Schmidt 2008). Coordinative discourse refers to communications between policy actors involved in policy construction, such as civil servants, elected officials, industry, and non-governmental organisations. Communicative discourse refers to communications between political actors – political leaders, spokespersons, members of the opposition, and activists – and the public, often about the necessity or appropriateness of proposed policies (Schmidt 2008). It is important to note that certain actors may operate within both the political sphere and policy sphere. For example, elected officials could provide their input during stages of policy construction and subsequently communicate their opposition or support for the constructed policy, based on the inclusion of their input in the eventual outcome. Most researchers nevertheless tend to focus on just one particular part of the public sphere (i.e. policy sphere or political sphere) and therefore analyse either communicative or coordinative discourse (Schmidt 2008). This thesis closely examines both coordinative and communicative forms of discourse, as the prospect of policy change is communicated both to policy actors and the public.

Furthermore, discourse should not only be considered a tool for communicating (calls for) policy change, but also an institution in and of itself that has the potential to affect policy change. This ‘discursive institutionalism’ entails that “policy discourses are used as guides to action by defining the concepts and norms to be applied, identifying the problems to be solved, developing the policy instruments to be used and framing the national policy discussion within a given policy arena” (Barbieri 2015). This theory is most recent and fourth addition to the study of ‘new institutionalism’ – distinct from rational choice institutionalism, historical institutionalism, and sociological institutionalism (Schmidt 2008). Discursive institutionalism maintains that understanding discourse helps explain the actions that lead to policy change, as the discourses of key policy actors may limit or increase the agency of decision-makers. In other words, analysis of policy discourse not only reveals whether policy windows are open and political opportunity structures are receptive to policy change, but also the interactive process in which different policy actors influence and develop the solutions presented by policy entrepreneurs, so that their policy preferences are incorporated within the actual policy change that is brought about by decision-makers.

(18)

17 3.4 Frame alignment

As framing is reflective of different policy actors’ perspectives, it can be used to discover the role of discursive institutionalism on the policy change process. This thesis therefore examines UK policy discourse on illegal and harmful online content using framing analysis to explain the UK government’s policy change. Framing generally refers to “the process of constructing and representing our interpretations of the world around us” (Gray 2003: 12). These interpretations concern “what the conflict is about, why it is occurring, the motivations of the parties involved, and how the conflict should be settled” (Gray 2003: 12). As framing takes place through discourse and implies specific policy solutions, framing analysis therefore reveals how discourse affects the actions of decision-makers and is a suitable approach for explaining policy change. Framing analysis and frame analysis are often confused as the concepts of ‘frame’ and ‘framing’ are closely related. Essentially, framing describes how actors use frames. A frame refers to “an account of ordering that (…) that describes the move from diffuse worries to actionable beliefs” (Hajer & Laws 2006). Frame analysis usually involves examining texts to identify pre-determined, generic or issue-specific frames. As framing analysis allows for “a more process-oriented and politically sensitive understanding” (van Hulst & Yanow 2016: p. 92) that changes over time, this is a more suitable approach for discursive institutionalist studies, which takes into account the development of policy discourse over time.

In particular, the concept of ‘frame alignment’ is central to the framing analysis and operationalisation of discursive institutionalism in this setting. Frame alignment describes the interactive process of alignment between policy frames – organising principles that reconstruct dispersed information into coherent, meaningful policy problems and (implied) solutions (Verloo 2005; Fletcher 2009; Barbieri 2015) – and a ‘master frame’. When policy actors’ framing of a policy topic demonstrates different conceptions of what the problem is and how it can be solved, this means that they are employing different policy frames. A master frame refers to “meaning on a broad scope and organizes sets of domain-specific frames” (Snow et al. 1986). Master frames are relatively stable and provide “a principal and official focus of attention for actors in the policy process: an arena of policy legitimacy and continuity” (Barbieri 2015). As “policy change is conditioned by the variations in the process of symbolic legitimation of government intervention” (Barbieri 2015) and frame alignment explains support and calls for different levels of government intervention, frame alignment can be considered to condition policy change. In particular, it provides insights into actors’ agency and how this affects the policy change process (Barbieri 2015). Frame alignment comes in two forms: frame amplification and frame extension. While frame amplification refers to the use of policy frames to clarify or

(19)

18 invigorate the master frame, frame extension refers to the use of policy frames to extend the boundaries of the master frame in order to incorporate different actors’ perspectives (Barbieri 2015). As such, studying discourse on illegal and harmful online content in the UK and determining frame alignment with the master frame, both before and after the policy change took place, will help explain the policy change.

It is important to note that policy actors can use ‘reframing’ deliberately to increase frame alignment with the master frame and encourage policy change. Reframing refers to the process of developing “a new way of interpreting or understanding the issues in the dispute or a new way of appraising one or more of the other parties in the conflict” (Gray 2003: 32). This deliberate use of framing can therefore be considered a political strategy, to influence the perception of others and mobilise support on a policy topic (Meyer & Staggenborg 1996). So, when does reframing reflect a change of mind and when is it merely a strategy to generate support for policy continuity or change? Answering this question, requires an interpretative approach that takes into account context and specifically, frame-makers’ identities and relationships with other stakeholders (van Hulst & Yanow 2016). This thesis therefore sets out different stakeholders and their policy preferences in terms of the regulation of platform liability in Chapter 5.

3.5 From theory to analysis

This theoretical framework forms the groundwork of this thesis’ analysis. The analysis is structured into three parts based on the three subquestions set out in section 3.1. Part 1 (Chapter 5 & 6) is largely descriptive, setting out the general policy preferences of key policy actors and determining policy change in the UK’s approach to the regulation of online platforms and specifically their intermediary liability for illegal and harmful content. This is necessary in order to explain how the policy change came about. This involves (1) identifying key policy actors and their policy preferences; (2) determining whether, how, and what point the government’s policy pathway shifted from policy continuity to policy change; and (3) discussing how the EU’s intermediary liability regime may have limited the scope for policy change by prohibiting certain forms of regulation and how likely it is that Brexit will eventually rid the UK of this limitation. Part 2 (Chapter 6) forms the core of the analysis, establishing how different policy actors employed framing within UK policy discourse on illegal and harmful online content and the extent to which their policy frames aligned with the identified master frame. This ‘frame alignment’ is key to explaining the agency the government had in changing its regulatory approach and bringing about this policy change. In doing so, this part considers (1) how key

(20)

19 policy actors (strategically) aligned their policy frames with the master frame and whether their policy frames highlighted the need for policy continuity or policy change through further government intervention, and (2) how this may have restricted the government’s room for manoeuvre and influenced the government’s approach to the regulation of platform liability. Finally, Part 3 (Chapter 7) considers other explanations for policy change – i.e. the actions of other states, government change, and focusing events – to contextualise the role of policy discourse on the government’s policy change.

4. Methods

This chapter explains the methodological foundations of the analysis in this thesis. It first describes the design of its analysis and data collection strategy, next provides justification for performing single sample case study of the UK’s intermediary liability regime and policy discourse, and finally explains the concrete steps that the framing analysis will take to accurately determine the legitimation of policy change.

4.1 Research design

This thesis operationalises the research question and subquestions using a qualitative methodological approach. This research contains both descriptive and explanatory parts. It first sets out the (policy preferences of) key actors in the UK platform economy – those which are frequent contributors to both communicative and coordinative policy discourse on illegal and harmful content in the UK: HM government, major political parties, industry representatives, and non-governmental organisations. It next discusses the development of the government’s approach to online platforms and identifies where this moved from policy continuity to policy change, as well as explaining the role of the EU and Brexit on UK policy change regarding platform liability regulation. The framing analysis itself covers both coordinative and communicative UK policy discourse on illegal and harmful online content, from when David Cameron started his second ministry in May 2015 until May 2018. As such, this framing analysis covers exactly three years of policy discourse and encompasses four important time periods: David Cameron’s second ministry, Theresa May’s first ministry, the 2017 general election campaign, and her second ministry (so far). Finally, it briefly considers alternative explanations for policy change, such as the actions of other states, government change, and focusing events. In terms of data collection, this thesis triangulates data from policy documents, news media articles, and academic literature to determine and contextualise the UK’s policy change. Data triangulation entails using data from two or more types of sources to validate findings

(21)

20 (Bogdan & Biklen 2006). More specifically, it analysed UK policy documents to determine and explain the policy change, EU policy documents to determine limits to the scope of national policy change, and policy documents of other Member States to provide further context to the UK’s policy change. News media articles and academic literature included in the analysis serve to provide further commentary on regulatory developments and to determine the policy positions of non-governmental stakeholders. For the framing analysis, this thesis collects data from both coordinative and communicative policy discourse – see Annex A for the specific list of materials. Consulted coordinative policy discourse includes parliamentary debates, policy consultation contributions, and official letters. Communicative discourse includes legislation, official statements, reports, publications of political parties (i.e. election material), and speeches. In order to provide a comprehensive account of framing in UK policy discourse, it analyses approximately 700 pages of UK policy discourse on illegal and harmful content (see table 1 & 2). In doing so, the framing analyses seeks to balance between communicative and coordinative policy discourse, contributions from policy actors, and pages per time period as much as possible.

Communicative

discourse Coordinative discourse Total

HM Government 169 68 236

Political Parties 59 125 184

Online platforms & industry 74 67 141

NGOs 51 88 139

Total 353 348 701

Table 1: total number of pages of policy discourse per key policy actor

Communicative discourse

Coordinative discourse

Total

Cameron’s first ministry 66 88 154

May’s first ministry 60 82 142

2017 general election 70 50 120

May’s second ministry 157 128 285

Total 353 348 701

Table 2: total number of pages of policy discourse per time period of the case study

4.2 Case study

As this thesis focuses almost exclusively on the UK, it can be considered a comprehensive case study of the UK’s intermediary liability regime and UK policy discourse on illegal and harmful online content between May 2015 and May 2018. The reason for selecting this as the starting point of the framing analysis is that policy discourse under Theresa May’s predecessor, David Cameron, during his last ministry may provide some explanation for the policy change brought

(22)

21 about by Theresa May. Single sample case studies are often avoided as they cannot be used to generalise theories (King, Keohane & Verba 1994; Hancké 2009). However, “case studies are not made for generalisations, should therefore not be used for that purpose, and (…) leaving behind the idea that single cases allow you to generalise opens a door to a much richer (…) way of using case studies” (Hancké 2009: 61). This thesis does not intend to generalise its findings across Member States but rather seeks to explain this example of relatively unexpected policy change. However, this does not mean that the UK cannot learn from other Member States and vice-versa. In fact, this thesis complements the findings of the framing analysis by examining the actions of other states and determining whether the UK is likely to have drawn inspiration from their approaches. In addition, other Member States may have followed or may still follow in the UK’s footsteps.

4.3 Framing analysis

In order to operationalise the research questions, this thesis employs an interpretative and bottom-up approach to framing analysis. Firstly, it selects policy discourse on illegal and harmful online content – policy documents where the topic of illegal and harmful online content features significantly – between the start of David Cameron’s second ministry in May 2015 until May 2018. Secondly, it scours through these documents in order to identify a master frame or different master frames. Thirdly, it determines patterns within the policy discourse, identifying these as policy frames and codifying three aspects of policy frames: “the construction of the public problems (…) (diagnosis), the problem-solving measures to be taken (prognosis) and the interaction patterns between the different agents (motivation)” (Barbieri 2015). In this, the motivation dimension refers to actors’ capability to reconstitute role attribution within policy alternatives (Barbieri 2015). Fourthly, this thesis discusses these policy frames in relation to generic frames that have been established in academic literature, as this helps explain whether policy frames directly conflict with one another or merely highlight different aspects. Finally, it determines how these policy frames align with the identified master frame in order to better understand the UK government’s policy change. In doing so, this thesis considers whether and how reframing is used deliberately to influence public opinion in light of the context in which statements were made.

4.4 Profiles: key policy actors

This section sets out the key policy actors in the UK platform economy and their policy preferences in terms of online platforms’ intermediary liability for illegal and harmful content.

(23)

22 Gaining a better understanding of the policy actors involved in the development of the UK intermediary liability policy regime is crucial to the interpretative and bottom-up approach of the framing analysis. This allows this thesis to consider policy developments from different policy actors’ perspectives and scrutinise their use of framing.

HM Government

The UK government is mandated to introduce national intermediary liability policies and is responsible for the transposition and implementation of EU-level policies. Key government departments of HM Government in this regard include the Prime Minister’s Office, DCMS, and the Home Office. High-profile members of the cabinet during the course of the case study include the prime ministers (David Cameron and Theresa May), the culture secretaries (Ed Vaizey, Karen Bradley, and Matt Hancock), and the home secretaries (Theresa May, Amber Rudd, and Sajid David) between May 2015 – May 2018. Furthermore, Baroness Joana Shields held important roles within the government as Minister for Internet Safety and Security from May 2015 to June 2017 and the Prime Minister’s Special Representative on Internet Safety from September 2017 to February 2018. It is in the interest of the entire UK government to implement a balanced intermediary liability regime: one that stimulates the development of online platforms founded in the UK, attracts inward investment and the establishment of foreign platform companies, limits disruption for British industry incumbents, and protects British citizens against illegal harmful content.

Major UK political parties

Alongside the ruling Conservative Party, the Labour Party played an important part in UK policy discourse on illegal and harmful content as the opposition between May 2015 – May 2018. Together these two parties had 87% of the total seats in Parliament during David Cameron’s second ministry and first ministry, and have had 86% of the total seats since the start of Theresa May’s second ministry. Relevant members of the Labour Party’s frontbench include the Leader of the Opposition (Jeremy Corbyn), the shadow culture secretaries (Maria Eagle, Kelvin Hopkins, and Tom Watson) and the shadow home secretaries (Andy Burnham and Diane Abbott). Other influential politicians in terms of platform liability include both parties’ backbenches in the House of Commons, members of the House of Commons DCMS Select Committee, and members of the all-party parliamentary group (APPG) on Intellectual Property and APPG Internet, Communications and Technology. As both parties are also represented in the House of Lords, the House of Lords Communications Select Committee and European

(24)

23 Union Select Committee also have an influence on UK policy discourse. It is clearly in the interest of Conservative Party MPs to support the government’s policy position, especially since its coalition with the DUP, and it is generally in the Labour Party’s to oppose the position of government, as they typically support further government intervention than the Conservative Party and seek to destabilise the government to gain political and public support. Some MPs or members of the House of Lords may nevertheless oppose the position of their parties, based on their own principles or those of their constituencies, on the topic of platform liability.

Online platforms and digital trade associations

Online platforms are of course at the heart of intermediary liability in the UK platform economy. They drive innovation and find more efficient ways to connect consumers, businesses, and citizens. While larger platforms have entire departments devoted to lobbying British and European institutions, others are represented on the policy stage through industry associations and membership bodies, such as the Confederation of British Industry, techUK, and the Internet Service Providers Association (ISPA) United Kingdom. It is in the interest of online intermediary platforms to inform governments of their level of passiveness in connecting users and to ensure that intermediary liability regimes do not become stricter. This would require them to become more actively engaged in monitoring their users content, placing a disproportionate burden on them and hampering the development of online platforms. Furthermore, such measures particularly affect smaller online platforms and start-ups as they do not have the same resources as established, international online platforms.

NGOs

NGOs are not directly affected by intermediary liability policy regimes. Nevertheless, they play an important role in shaping attitudes towards online platforms through their campaigns and research. As the voice of civil society, NGOs often represent both commercial and general users of online platforms. Commercial users of online platforms take various forms. Artistic entrepreneurs may use online platforms like Etsy to sell their photographs or graphic designs, scholars can publish their articles on SSRN or Academia to share their insights with the world, and universities can offer online courses on Coursera. As online platforms are also a breeding ground for violations of IP rights, it is in the interest of commercial users that online platforms take down any content that violates copyright law. General users of online platforms may be active in an economic capacity as consumers, in a social capacity connecting with friends or professional contacts, or to obtain information as inquisitive citizens or researchers. However,

(25)

24 users may be affected by other users that abuse platforms to bully, ‘troll’, intimidate, offend or harass others. Furthermore, certain platforms are used by extremists to distribute propaganda and criminals or terrorists to communicate with one another. It is therefore in the interest of the majority of general users that these forms of illegal and harmful content are taken down and that platform-abusing users face repercussions. Example of a British NGOs that advocate this position are the Internet Watch Foundation (IWF), the Children's Charities' Coalition on Internet Safety, the UK Safer Internet Centre, and Childnet International. Overall, NGOs take the position that online platforms need to act (more) responsibly towards their users.

5. Determining policy change

When explaining Theresa May’s reforms to the UK’s intermediary policy regime, this thesis must confirm that this indeed constitutes ‘policy change’, but also determine the kind of policy change it constitutes and how it came about. The first section of this chapter examines the development of the UK intermediary liability policy regime and identifies the exact moment(s) of policy change. It examines the regulatory approach the UK government took towards the intermediary liability of online platforms for illegal and harmful content: self-regulatory, co-regulatory, or statutory, which allows this thesis to quickly identify whether policy change concerned core or secondary aspects – the government’s deep core policy approach to government intervention remained unchanged during the case study under consecutive Conservative Party governments. Furthermore, examining the development of the UK’s approach to platform liability for illegal and harmful content provides useful context for determining the significance and magnitude of policy change. The second section considers whether the EU’s intermediary liability policy regime constrained the UK’s room for manoeuvre in terms of its regulatory approach and discusses the likely effect of Brexit on this regulatory relationship. Determining EU policy constraints is essential to understanding the type of policy change the UK government has brought about and intends to post-Brexit, as EU rules may have prohibited certain regulation and thus limited the scope for national policy change.

5.1 Development of the UK intermediary liability policy regime

As policy change can only be determined in light of an original policy designs this section first describes the UK’s regulatory approach to platform liability from its first policy measures in 1996 until David Cameron became prime minister in 2010. It next examines policy developments throughout David Cameron’s tenure in order to determine the platform liability status-quo – the evolved design of the UK’s intermediary liability policy regime – before Theresa May succeeded