Lots has been written about the epidemiological modelling of the Coronavirus outbreak but I’ve seen very little published about any human behaviour modelling. I’d be interested to know what models have been created to describe how people might behave in different pandemic scenarios. Here, I want to concentrate on some ways in which people’s behaviour can be (is being) shaped to help in the fight against Covid-19.
My Christmas reading was “Thinking, fast and slow” by Daniel Kahneman and I’m still reading it (it’s not an easy read). Kahneman is a psychologist and economist who won a Nobel Prize for his work on behavioural economics in 2002. Thinking fast and slow is all about why people think what they do and why they make the decisions they make. Kahneman calls “thinking fast”, System 1, and “thinking slow”, System 2. System 1 operates automatically and quickly, with little effort and without voluntary control. System 2 requires mental effort and concentration.
For example, if you were asked what is 2 + 5 you would instantly know that the answer is 7. System 1 gives us the answer automatically because it’s something we’ve learned through practice and repetition. If, however, I asked you to multiply 17 x 23, you’d probably have to think more carefully and get out a calculator. However, System 1 would probably tell you that a guess of 200 would be too low and 500 would be too high. Kahneman claims that we all like to believe we think in a System 2 way; i.e. rationally, and we use that ability to make informed decisions. Unfortunately, that’s not what happens in practice.
System 1 is fine because it uses what we have learned in order to react quickly and lowers the mental load we have to cope with. System 1 often uses “rules of thumb” to make decisions quickly and usually these lead to good decisions.
Chapter 30 (!) deals with “Rare Events” and focuses on our inability to understand statistics and probabilities. It is very relevant in relation to changing people’s behaviours to respond to the Covid-19 pandemic. In summary:
- People overestimate the probabilities of unlikely events
- People overweight unlikely events in their decisions
Although overestimates and overweighting are different ways that people react to data, they both result from 2 mechanisms called confirmation bias and cognitive ease. System 1 takes shortcuts to make decisions; for example, confirmation bias means you tend to agree with information which supports something you already believe. Cognitive ease refers to how easy it is for our brains to process something.
Vivid Probabilities make a bigger impact on our decisions
Kahneman also talks about “Vivid Probabilities” by which he means that the ease of imagining a particular outcome contributes to how people make decisions.
The publication of daily deaths from Covid-19 and charts showing the exponential rise in diagnosed cases plays to this concept of vivid probabilities. It also makes use of what has been called “denominator neglect” to trigger a weighted decision response in people.
If you read that 0.002% of the population had died from Covid-19, your reaction would be very different to seeing a statement that 1000 people had died. Reporting the number of deaths makes much more of an impact on people and, in the case of Covid-19, is more likely to make them adopt the actions the government wants them to. People can readily visualise individuals dying, whereas they don’t interpret percentages with the same vivid perspective. They also overweight the likelihood of them being one of the people who dies.
In one study, people who saw information that “a disease kills 1286 people in every 10,000” judged it as more dangerous than people who were told about “a disease that kills 24.14% of the population”. The first disease is perceived as more threatening than the second, even though the former is half the risk of the latter. This is also known as the denominator effect whereby different ways of expressing data vary so much in the impact they have.
The Daily Telegraph reported (today) that the current Covid-19 death rate varies between 0.7% and 3.4% (depending on the location and access to good hospital care). They didn’t report that between 966 and 993 out of every 1000 people will survive.
Availability bias and the availability cascade
Kahneman also discusses the “availability bias”, where you overestimate the probability of something that you have heard often or that you find easy to remember. The fact that Covid-19 deaths are reported daily (and discussed almost continuously in the media) means people overestimate the probability of them also dying. As a consequence, they are more likely to follow the advice on social distancing or self-isolation.
The “availability cascade” is a self-sustaining chain of events that may start with a few media reports of a problem that lead to widespread public panic and eventually result in policy changes by legislators. Often, the emotional reaction (e.g. people dying) becomes a story in itself and the story can be accelerated by media headlines, social media groups and campaigning individuals who work to ensure a continuous supply of bad news cases. It is certainly true that this has happened with Covid-19. The danger is that politicians and legislators enact policy changes that are a response to the panic rather than a response to the evidence. No doubt, once we have returned to whatever “normal” looks like, there will be analyses of the timing and content of the policies that have been implemented during the Covid-19 pandemic.
One other element that has been notable by its absence throughout most of the communication on the pandemic is what is known as “Reference Class Forecasting”. This is the use of comparative data to provide an outside view. In the world of project management, Bent Flyberg uses this approach to challenge optimism bias in project budgets and plans. By comparing a current project with numerous similar ones, it is possible to generate a better baseline prediction for a project plan. There has been some reporting of the scale and impact of other pandemics (e.g. SARS, MERS, N1H1, Ebola) and those can provide a useful reference to help people cope with the current uncertainties of Covid-19.
We might also have seen other reference data published, for example deaths from winter flu that occurs every year, because comparisons have been made with that disease. The UK government’s published flu data for the past 5 years is shown below.
From this, we probably shouldn’t be surprised to keep hearing that the elderly (who are likely to have more underlying health conditions) are more at risk of dying from Covid-19 than the young. Equally, we shouldn’t be surprised to hear reports of young people dying from Covid-19. The fact that the press makes big stories out of individual cases is another example of how “vivid probabilities” can be (are being) used to influence people’s behaviours in the fight to minimise the impacts of Covid-19.
We need System 1 and System 2 to fight Covid-19
I hope that there’s enough System 2 thinking going on behind the scenes and that a whole systems perspective is being taken. We ought not to be making policy decisions on the basis of models that focus narrowly on health (or deaths). We are certainly seeing how the UK government is relying on its senior scientific advisors both to inform policy and to communicate it.
It appears that an understanding of the differences between System 1 thinking (fast) and System 2 thinking (slow) is behind the way governments are communicating their policies to deal with Covid-19. The way that rare events are communicated and then overweighted by people is useful in nudging the right behaviours that will help keep us all safe.
Useful Covid-19 data and links (with thanks to Nigel Marriott).