Building trust in the digital era: achieving Scotland's aspirations as an ethical digital nation: case study supplement

Public Awareness of Data Use and Sharing

Case Study: Targeted Advertising, Advanced Marketing and Behaviour Change – Dr. Ben Collier & Dr. James Stewart

Contemporary forms of digital marketing are the financial lifeblood of the Internet. Most of the online platforms, search engines and social media sites we use are provided for free to the end user, generating revenue through the collection of intimate behavioural data, which are used to generate advertising profiles. These profiles allow adverts to be targeted and personalised not only based on demographic characteristics and traditional segmentation, but on previous and current behaviour, surfaced by the application of algorithmic technologies to extremely intimate and fine detail records of online browsing, communication, and activity. The targeted digital advertising industry has been the subject of a series of scandals and critical debates in recent years, not only due to concerns around intrusive corporate surveillance, but also in the use of this surveillance influence infrastructure for legitimate and subversive political communication. We have recently identified a new area of potential concern: the increasing use of these infrastructures by government to shape the behaviour of the public.

Government communication practices are not static, and change and adapt in line with the cutting edge of industry practice. These practices involve not only classic forms of awareness-raising - public health and safety, regulatory changes, democratic participation etc, but attempts to directly shape the behaviour of the public - often through ‘nudge’ and other approaches incorporating insights from behavioural science. As digital marketing tools have evolved, government departments and law enforcement are increasingly using them in behaviour change campaigns as part of a shift to prevention. This in theory allows government to shape behaviour in-the-moment in novel, intimate, and deeply targeted ways, bringing together administrative data, marketing data, and platform targeting data to target, deliver, and evaluate complex campaigns.

These tools complicate the ‘participatory’ and democratic nature of modern communications and government policy and raise some profound ethical issues. First, there are a series of practical and legal questions. The use of government administrative or survey data to develop targeting profiles may be contested where those data are explicitly not to be used for marketing purposes. This blurs the line between marketing and service delivery. Secondly, the algorithmic targeting of adverts leaves open legal room to challenge if it can be proven that it has the potential to harm or disadvantage.

The public are largely aware of the existence of digital targeting, and as a result, may feel anxious if they receive government adverts, which they assume are as a result of their online behaviour. Particularly for more vulnerable groups, this presents a real capacity for unintended harm. Additionally, there are a set of issues around privacy/intrusiveness. These practices open up to government a new generation of detailed data sources that can be used to target communications by interposing a private entity (the platform). This allows for the use - at arm’s length - of very intimate targeting and delivery approaches in ways not historically available to government.

Within the Scottish Government, many of the policy campaigns are run with the Marketing, Insight, Brand Scotland and Internal Communications Unit, in particular the, Strategy & Insight Team, acting as gatekeeper, experts, and promoter of the design, targeting, ethical review, and purchase of targeted ads. This is partly historic due to the conventional (and now digital) advertising in the Scottish government being placed through framework agreements that are tendered for by media buyers every (4 years) and administered by the MIBSICU. This unit puts out a call twice a year for policy units to propose targeted campaigns, including digital components, and circulates examples of successful practice, awards won etc widely in government. Units propose projects, these are evaluated at ministerial level, and resources of the Unit allocated according to ministerial priorities. The Unit arranges tendering processes with creative agencies to co-design the campaigns.

Although this unit is staffed with thoughtful and skilled staff who are attentive to ethical issues, there is no formal ethics training provided for dealing with these influence methods - rather, this is a case of expert practitioners learning and sharing on the job. There is a distinctive Scottish approach developing for the delivery of these campaigns, which draws on GCS expertise without replicating the same frameworks; this does integrate some of the expertise of Behavioural Science but not in a systemic way. Media buying is arranged through two agencies who have a lot of input into this process. SG gives them the target audience and the desired behaviour change - then they draw on marketing data, such as from Yougov etc. to develop profiles, then they do the ad buy.

We can broadly counter pose two distinct visions of government digital behaviour change - these are ideal types, intended to demonstrate potential positive and negative futures rather than referring to specific current practices. The first, we describe as ‘influence government’, embodying the more coercive and centralised forms of behavioural marketing. In this format, efficiency of delivery and central control of goals and resources are paramount. Campaigns are designed and delivered centrally, with little connection to local experiences and priorities. They are intrusive in the data, which are used to target, including a range of commercial sources which people would not expect to be available to government and a range of government sources people would not expect to be used for marketing. They are also intrusive in the spaces and places in which messaging is delivered, such as in the home environment or more intimate online channels. In these campaigns, complex policy issues with structural causes are simplified, with policy levers, and responsibility for social change, collapsed down to the level of individuals making choices. They are coercive and covert, framed as more successful when the individual is unaware, passively receiving messages which shape their behaviour.

The second vision might better be described as ‘participatory prevention’. This would draw on the Scottish Government’s core strategies around participation and co-production, following the lead of best practice examples in industry and public service where the public are not only consulted to develop the aesthetic wrapper of the message, or studied to inform targeting, but also directly involved in shaping the priorities, theories, and messages, which animate these campaigns. These approaches aim to empower, framed as more successful when people know that they are being messaged and why - they are actively engaged.

Recommendations

At its current level of development, it is clear that despite individual concerns about particular campaigns, there is the capacity for digital behaviour change to campaigns to be assessed by this central unit of practitioners. The expertise and responsibility is distributed between this central unit and key external suppliers - we are yet to explore the role of these suppliers in depth, however this clearly needs to be better and more systematically understood by government, as at present this appears to be somewhat of a black box. If the use of these practices were to be expanded, there is a further need to provide wider frameworks of accountability and review. The scaling up of these methods would likely require the creation of a professional oversight board within the Scottish Government, which would review proposals for new digital behaviour change campaigns by Scottish Government departments and agencies. Membership of this board could include communications professionals, academic experts, statisticians, and experts in data ethics. This board could meet fortnightly, modelled after the statistical corps’ data sharing board, developing expertise that can be re-used consistently and identify patterns. Legal and ethical issues with these approaches should be explored in further public debate about the use of these methods. More broadly, there is an immediate need for a wider public discussion of these issues - and an explicit determination at the level of policy and legislation of the acceptable bounds of such approaches.