The problem of digital self-control

In the ancient Greek epic poem The Odyssey, the hero Ulysses and his men must sail close by the Sirens, whose enchanting singing fatally distracts anyone who hears it. To overcome this self-control challenge, Ulysses stops his men’s ears with wax, commands them to tie him to the mast, and also to ignore him if, upon hearing the Sirens’ song, he asks to be untied. As they sail past, and the Sirens call out his name, Ulysses is overpowered by a desire to get close to them and signals to his men that they should untie him. His men, however, follow his earlier command and refuse to untie him, thereby holding him accountable to his past goal. Not until they are safely out of hearing range of the Sirens’ voices do the men take the wax out of their ears and unbind Ulysses.

This story is sometimes referenced by self-control researchers to illustrate one possible strategy — ‘pre-commitment’ (Elster 1984; Willigenburg and Delaere 2005) — by which people help themselves act in accordance with their longer-term goals in the face of temptation. Today, the advent of smartphones, lightweight laptops, and smartwatches has meant that most people keep a ‘Siren song’ from unwanted distractions permanently within reach. Thus, whereas the instant access to information, entertainment, and connection provided by such devices brings numerous benefits, many people also find that it makes it difficult for them to focus on their current task without being distracted by notifications, habitual check-ins, or endless opportunities for procrastination (Dabbish, Mark, and González 2011; Marotta and Acquisti 2017; Stothart, Mitchum, and Yehnert 2015; Thomas, Azmitia, and Whittaker 2016). A viral blog post summarised this as “I can’t handle infinity in my pocket” (Knapp 2013). The challenge is exacerbated by tech companies whose business models depend on getting people to use digital services frequently and extensively (cf. the ‘attention economy,’ Davenport and Beck (2001), Wu (2016); and ‘surveillance capitalism,’ Zuboff (2015)). This incentivises software designers to make users ‘hooked’ on their products and craft a Siren song that is as enchanting as possible (cf. Eyal 2014).

Catering to users struggling to control their use of digital devices, a niche has emerged for apps and browser extensions that provide a wide range of self-control interventions (‘digital self-control tools’), such as blocking access to distractions—and sometimes going to lengths to make blocking impossible to override, akin to Ulysses requesting to be tied to the mast—or more subtle approaches such as visualising time spent or rewarding intended use (Lyngs et al. 2019). Following growing public concerns over addictive design patterns (e.g., Harris 2016; Foer 2016; Tiku 2018), Apple, Google, and Facebook recently begun to implement similar ‘digital wellbeing’ features into their products (Apple 2018; Google 2018; Ranadive and Ginsberg 2018).

Meanwhile, a growing body of work in Human-Computer Interaction (HCI) is building an evidence base on what design patterns are actually effective in supporting users in exerting self-control over digital device use. Yet, this research is in its early stages. The existing studies provide initial indicators of how design patterns ranging from visualisations of device use (Whittaker et al. 2016) to goal-setting with social support (Ko et al. 2015) can be helpful, but we still lack high-quality evidence on basic questions: In what contexts are different design patterns more useful? To what extent does their effectiveness depend on individual differences? What does the overall design space look like for design patterns for digital self-control? Moreover, basic conceptual questions, such as how to define ‘success’ for these patterns, and which theoretical frameworks are most suited to guide research efforts, remain open. The work presented in this DPhil thesis aims to move research forward on these questions.

Research question, approach, and scope

Based on the existing literature, which is laid out in greater detail in Chapter 2, the thesis aims to answer the following research question:

How can existing digital self-control tools help us identify effective design patterns for supporting self-control over digital device use?

To answer this question, the thesis sets the following objectives:

• Explore how established self-regulation research can help us organise and evaluate existing design patterns, and provide a deeper understanding of the self-control struggles they seek to address (Chapter 3)
• Understand, at scale, how current digital self-control tools in online stores have explored the design space and how users have responded in store ratings and reviews (Chapters 3 and 4)
• Explore how digital self-control tools in online stores can be used to identify specific, promising interventions, and inform targeted studies (Chapter 5)
• Explore how the landscape of existing tools can help us elicit personal needs for digital self-control interventions in specific populations (Chapter 6)

In addressing these objectives, we narrow the scope of our investigations in the following ways (cf. Chapter 7):

• we focus on design patterns for digital self-control on smartphones and laptops,
• we focus on individual users’ ability to exercise self-control (as opposed to, e.g., how system design impacts communities at a group level, cf. Hekler et al. (2013)),
• we exclude design patterns for regulating use in relation to families, children and young teenagers, as well as distracted driving, because design motivations in these cases include wider normative questions over, e.g., promotion of ‘healthy’ child development, or policy or legal concerns over traffic safety, and
• we focus on design patterns motivated by self-control struggles experienced by average users, as opposed to more clinical cases of severe ‘addictive’ use, where the underlying main causes are likely to be unfulfilled psychological needs in the user’s life more broadly rather than digital devices’ design affordances per se (Przybylski and Weinstein 2019).

Finally, in terms of research approach, we address our objectives through a combination of quantitative and qualitative methods, in addition to reviewing existing literature:

• we review cognitive neuroscience literature on dual systems approaches to self-regulation and contextualise it to digital self-control (Chapter 3),
• we combine web scraping, feature coding, and thematic analysis to characterise the existing landscape of apps and browser extensions for digital self-control (Chapters 3 and 4),
• we combine usage logging, surveys, and semi-structured interviews in a longitudinal study to evaluate specific, promising interventions identified among existing tools (Chapter 5)
• we collaborate with the University of Oxford Counselling Service on a workshop and intervention format to help understand how current interventions may address digital self-control struggles among students (Chapter 6).

Contributions

The thesis provides a number of contributions to research on digital self-control:

An extension of current applications of dual systems theory in HCI, that is in sync with basic research on self-regulation. The dual systems framework presented in this thesis incorporates the ‘expected value of control,’ which in the recent cognitive neuroscience literature is considered central in explaining why success at self-control fluctuates over time and with emotional state. This demystifies the concept of self-control and helps clarify how specific design features may work to scaffold successful self-control.

The first systematic analysis of digital self-control tools on the Google Play, Chrome Web, and Apple App stores. While a couple of existing papers have investigated this landscape, the number of tools studied in this thesis is approximately an order of magnitude larger than previous studies, and also provides the first analysis of user numbers and ratings. This contributes a broad characterisation of how the design space have been explored by current tools, as well as an indication of the effectiveness of specific design patterns and implementations.

The first demonstration of how design interventions can support self-control on Facebook. Whereas a large number of studies have investigated ‘Problematic Facebook Use,’ a very small number have used experimental methods, and none have investigated the effectiveness of interventions common in online stores for browser extensions. This thesis demonstrates that two specific interventions, drawn from the Chrome Web store, can potently influence behaviour and reduce unintended use, which suggests potential remedies for problematic use that present a less tall barrier to action than deactivating or deleting one’s account.

A new workshop format for in-depth studies of personal digital self-control struggles and appropriate interventions among university students. While numerous studies have highlighted these struggles among students, the present thesis work is the first to focus on this demographic in a workshop that embeds interventions drawn from comprehensive reviews of digital self-control tools. The thesis contributes open materials for the workshop alongside implications for how design patterns may be better aligned with users’ goals via focused, within-app interventions, or by supporting the formation of habits and beliefs that help users stay in control with less reliance on external interventions.

An advancement of open and transparent research in the field. Open science practices have been called for in HCI (Haroz 2019), but has seen little adoption in digital self-control research. This thesis represents the first work in the field to make its materials, data, and analysis pipelines openly available, as well as to use a reproducible plain text format (R Markdown, Xie, Allaire, and Grolemund (2018)) for the manuscript source files.¹ Moreover, the thesis work has led to the creation of a number of R packages for reproducible academic writing (e.g., for ACM CHI submissions), all of which are openly available.

Terminology

Throughout this thesis, we will use the terms

• ‘digital self-control’ to describe users’ ability to align their digital device use with their valued, longer-term goals,
• ’design patterns’ to refer to specific functions intended to support user self-control (e.g., hiding newsfeeds, providing goal reminders),
• ’intervention’ to refer to design patterns when they have been instantiated in a piece of software (i.e. a digital self-control tool or feature) being evaluated, or when considering the way a design pattern changes a user’s usual digital environment,
• ‘digital self-control tools’ (DSCTs) to refer to apps and browser extensions that implement a given design pattern (e.g., a browser extension for hiding Facebook’s newsfeed), and
• ‘conceptual framework’ to refer to theories that provide an account of how fundamental building blocks, or ‘constructs,’ are interrelated in human behaviour at one or two distinct levels (following Hekler et al. (2013), see section 2.3.5.1).

Thesis outline

The structure of the thesis is as follows:

In Chapter 2, we summarise the background and motivation. We first review work on user self-control struggles and the challenges of defining ‘success.’ Afterwards, we review findings and theory applied in existing studies of design patterns for digital self-control, before summarising the current research limitations and opportunities that motivate this thesis.

In Chapters 3 and 4, we present a broad view of how existing tools can help us identify effective design patterns. Chapter 3 characterises the design space for digital self-control design patterns, in two ways: ‘bottom-up’ by analysing features of current digital self-control tools, and ‘top-down’ by applying a dual systems framework. To this end, we first formulate a dual systems framework of self-regulation, and contextualise it to digital self-control. Afterwards, we conduct a systematic search for apps and browser extensions for digital self-control on the Google Play, Chrome Web, and Apple App stores, categorise their features, and analyse them using the dual systems framework.

In Chapter 4, we extend Chapter 3’s feature review by collecting store popularity metrics and user reviews for the same set of digital self-control tools. We then conduct a combined analysis of tool functionality, user numbers, average ratings, and the content of reviews, and interpret the findings using the dual systems framework.

In Chapter 5, we illustrate how the landscape of existing digital self-control tools may be used to identify specific research opportunities and drive focused studies: we conduct the first controlled study of how user interface interventions on Facebook — drawn from popular extensions on the Chrome Web store — affect behaviour and perceived control among university students. We use the dual systems framework as a lens for selecting interventions, interpreting the findings, and predicting the limits and potential of the design patterns implemented by the interventions.

In Chapter 6, we show how the range of existing design patterns can be used in an active workshop to elicit user needs for digital self-control interventions: we report on four ‘Reducing Digital Distraction’ (ReDD) workshops conducted with students at the University of Oxford. We use the dual systems framework to analyse the interventions included, and to discuss workshop findings in relation to tools that serve as ‘training wheels’ for improving self-discipline over time.

In Chapter 7, we discuss contributions of the thesis as a whole, in addition to wider challenges and opportunities regarding methodology and application of theory in digital self-control research. The thesis closes by revisiting what ‘success’ might look like, before presenting a longer-term vision for the ReDD workshops as a tool for data collection in digital self-control research, as well as for practical impact.

Dissemination

This thesis’ research has been disseminated in the following papers, blog posts, and open source contributions: