Five ways behavioural insights can improve COVID tracking apps
A few weeks ago, the EU announced that the use of tracking apps for the ongoing pandemic would be voluntary. Best estimates say that at least 60% of people need to effectively use such an app for it to actually mitigate contagion (EU eHealth Network, 2020). So, I thought – wow! Are they really trusting me and all other Europeans to choose wisely about using the apps? The best uptakes we've seen so far are well below this rate (Iceland at 40% and Singapore at 20%. See the helpful summary by Thomas Pueyo, 2020 ). Individual decision making is affected by various biases inside our own minds – they're called cognitive biases – that prevent us from assessing problems through a fully-fledged cost-benefit analysis. For example, we don't rationally weigh all the public health risks of not using the apps vs. the actual risks of data breaches or the hassle of keeping up to date with the app. I wouldn't even know where to start!
As a behavioural economist, my first thought was that we'll soon be back in a full lock-down as we face slim chances to successfully achieve the necessary wide-spread use. Then my reasoning took a more constructive turn. I realised that behavioural science can help. Especially because a few features of these tracking apps (voluntary download, honest disclosure of symptoms, the public good conundrum etc.) are particularly suited for improvements based on behavioural science. This approach complements the one suggested by my colleagues in this recent piece on how seamless health ecosystems would improve the effectiveness of such apps.
This is a short overview of these behavioural insights, arranged by a few outcomes that are instrumental for the apps' effectiveness. These are my opinions, but based on the learnings from hundreds of trials that we've done at Swiss Re with insurers to improve various aspects of consumer engagement, including the take up of health behaviour change apps.
Five ways behavioural insights can improve the effectiveness of COVID tracking apps
The single most important measure is the first one on my list – opt-in defaults – because it creates the highest potential for impact.
1. Increase download and usage
Try different methods to ensure the public health benefits won't be discounted against people's overinflated perceptions of other risks.
- Using opt-in defaults has been a successful strategy for other public good challenges, such as organ donations (Johnson and Goldstein, 2003) and pension schemes (Choi et al, 2003). They've been able to lift take-up rates from approximately 15% to 99%. Would an automated opt-in scheme also be allowed for the download of such apps? This would imply that they would be automatically downloaded on one's smartphone, perhaps with the next software release, unless you opt-out – which would maintain one's free-will with regard to this choice. Thomas Pueyo is also in favour of this approach (Pueyo, 2020).
- Use the principle of reciprocity to inform download messages. i.e. ‘When you go out, would you like to be among healthy people? If so, please grant others the same security and download the app’. This has proven successful in other health-related interventions, such as to encourage the sign-up to the organ donation registry in the UK (Behavioural Insights Team, 2013). Another principle of relevance could be the ego effect, i.e. ‘Download the app to do the right thing and contribute to the wellbeing of your country’. This might counterbalance the users' perception that their short-term freedom to go out and be able to continue with their lives is the most beneficial choice – a risk that a number of behavioural scientists recently pointed out (Chater et al, 2020).
- Use enhanced active choice prompts when someone ignores the option to download the app, highlighting the costs of such a choice, i.e. ‘If you delete this app you may increase the possibility of the next lock-down. Are you sure you want to proceed? ’ Keller et al (2011) show how this is impactful to create healthy habits.
- Use social norms/curiosity messaging once take up starts and relevant data becomes available, i.e. ‘Would you like to know how many people in your area are using the app or are showing symptoms? Download the app’.
- Use the messenger effect for the invitation to download the app, so that it comes from a trusted party (i.e. the Prime Minister, the health system, etc.).
- Design the apps so that rewards (such as public recognition or a points system) are an integral part, as a pay-back if the app is shared with one's network. To encourage such behaviour further, reinforce the point through social norms/externalities messaging, i.e. 'Sharing the app will contribute to xyz'. There was promising evidence on this working well especially for men, from a recent online experiment with a sample of Italians by Jachimowicz and DiTommaso (2020).
- Use social norms and include top-tips for compliance in messages for those who don't comply. This has proven powerful to influence decision making to benefit the public in other public health settings, such as GP's prescription behaviour to mitigate the risks of antimicrobial resistance (Hallsworth et al, 2016).
- Ask users about their motivations for using the app, so that these could be used to inform a commitment device or reminder in case their app-use decreased.
2. Minimise the perceived and actual risks of data-privacy breaches
Importantly, the use of such apps should not just strive to maximise public health benefits, but also ensure that data privacy and cyber risks around data sharing are minimised. There are a few insights relevant to this aspect, which should be combined with effective technical solutions which are beyond the focus of this article:
- Clarify ‘why’ data is being sought
- Design them so that the default option (i.e. the do-nothing option) will ensure the highest probability of data-deletion
- Give control and make it easy for users to delete their data
- Use a very simple tick-box visualization to summarise terms and conditions around data security measures in place, to increase the users' understanding.
3. Increase the honest disclosure of symptoms
We at Swiss Re and several academics have done extensive research on honesty in disclosing health symptoms when incentives are misaligned, i.e. when being honest can generate some personal costs. Based on what we've learned, here are a few ideas for how to ask which symptoms someone is experiencing in a way that would encourage honest answers.
- Add honesty-prompts (such as asking for an honesty declaration and fingerprint-recognition upfront) when asking users to disclose their symptoms – there is mixed evidence on this approach but it could be effectively combined with the insights below.
- Externalities messages seem to be working especially for young people, i.e. 'Sharing the app will contribute to xyz' (Jachimowicz and DiTommaso, 2020).
- Don't ask people to calculate things, to avoid the fudge factor.
- Be wary that the answers to the questions at the end of a form may be less honest as a result of ‘moral self-licensing’ in late disclosures (Merritt et al, 2010).
- The ‘truth-serum’, according to which people project their behaviour onto their peers, implies that asking someone to disclose their symptoms and their beliefs about their peers' symptoms could shed light onto whether their answers about themselves were honest. Make sure to add these questions to the health disclosure forms (Prelec, 2014).
- Design questions so you don't make it obvious which one is the ‘freedom-limiting' answer, to prevent users from gaming the system.
- Where relevant, you could also explain that the accuracy of one's symptoms could be confirmed with their health practitioners; this would make the perception of possibly being checked upon more tangible (where this is realistic).
4. Encourage compliance with the relevant physical distancing rules
Inducing health behaviour change is hard! This has been studied extensively, in relation to weight loss, smoking cessation, medical adherence and many other aspects. Here are some behavioural insights informed by this body of evidence that could help encourage people to stay at home when they need to for everyone's health benefit.
- Use reciprocity and social norms messaging to encourage young people to stay at home (Jachimowicz and DiTommaso, 2020).
- Gamification (tiers of rewards), reinforcement-rewards (small, frequent and timely) and lotteries are all relevant systems to induce health behaviour change (Vlaev et al, 2019).
- Consider how to include functionality for peer advising by top performers to share how they're complying with the relevant physical distancing measures and the use of the app.
- Let users set up personal goals within the context of what the apps are trying to achieve, this would help users achieve them.
- Traffic-lighting (increasing the size of the red and orange ranges), instead of ‘technical definitions’ is a much better way to communicate to users about risks.
- Use growth-mindset types of encouragement. This is about encouraging people's efforts on their journey toward a certain goal rather than just when they achieve it (Dweck, 2008).
- Timeliness is key. Ask users to build new habits at a time like the one when they’ll feel the benefit 'stay at home if you have these symptoms so that later this spring you will be safe to go out' (vs. next month) or focus on scarcity ‘do this by x date’.
5. Help tracking apps succeed in other ways
Here is a final collection of insights that are equally important and apply across all categories:
- A/B test everything and understand for which group which of the above approaches works best and try and tailor the app accordingly.
- Adopt simple language, i.e. the reading age must be below grade 6 (i.e. an 11-year-old should understand it); this is actual guidance from the EU, most often failed by online websites as assessed by Brooks et al 2013. Use a warm tone matching people's expectation when discussing their health.
- Ensure that the healthy option (i.e. the relevant guidance to react to one's symptoms) is the path of least resistance, i.e. make it simple for users to follow it.
- Adopt as much personalisation in all communication as possible and visualisations to ensure that the risks implied by non-compliance to physical distancing rules are better understood.
We’ve gone so far already in terms of mitigating the impacts of the pandemic, and these many small – but important – changes could make a true difference. Let's make this next phase of pandemic management as behaviourally-informed as possible!
Behavioural Insights Team (2013). Applying behavioural insights to organ donation. Accessible here.
Brooks, C., Ballinger, C., Nutbeam, D., & Adams, J. (2013). Literacy levels required to understand regularly accessed falls prevention websites aimed at the public. Journal of Physical Therapy and Health Promotion, 1(1), 8-14.
Chater, N., Delaney, L., Dolan, P., Hahn, U., Lordan, G. (2020). Behavioural Science in the Context of Great Uncertainty. Wednesday 13 May 2020 2:00pm to 3:30pm. Hosted by LSE's public event series - COVID-19: The Policy Response. Accessible here.
Choi, J. J., Laibson, D., Madrian, B. C., & Metrick, A. (2003). Optimal defaults. American Economic Review, 93(2), 180-185.
Dweck, C. S. (2008) Mindsets: How Praise Is Harming Youth and What Can Be Done about It. Accessible here, pp 57-58.
EU eHealth Network (2020). Mobile applications to support contact tracing in the EU’s fight against COVID-19. Accessible here.
Hallsworth, M., Chadborn, T., Sallis, A., Sanders, M., Berry, D., Greaves, F., ... & Davies, S. C. (2016). Provision of social norm feedback to high prescribers of antibiotics in general practice: a pragmatic national randomised controlled trial. The Lancet, 387(10029), 1743-1752.
Jachimowicz, J. M., DiTommaso, D. (2020). Accessible here.
Johnson, E. J., & Goldstein, D. (2003). Do defaults save lives?.
Keller, Punam & Harlam, Bari & Loewenstein, George & Volpp, Kevin. (2011). Enhanced active choice: A new method to motivate behavior change. Journal of Consumer Psychology - J CONSUM PSYCHOL. 21. 376-383. 10.1016/j.jcps.2011.06.003.
Merritt, A. C., Effron, D. A., & Monin, B. (2010). Moral self‐licensing: When being good frees us to be bad. Social and personality psychology compass, 4(5), 344-357.
Prelec, D. (2004). A Bayesian truth serum for subjective data. science, 306(5695), 462-466.
Pueyo, T. (2020) Coronavirus: How to Do Testing and Contact Tracing. Part 3 of Coronavirus: Learning How to Dance. Accessible here.
Vlaev, I., King, D., Darzi, A., & Dolan, P. (2019). Changing health behaviors using financial incentives: a review from behavioral economics. BMC public health, 19(1), 1-9.