I was live tweeting throughout the Behavioural Exchange conference, which was held in Sydney. It brings together policy-makers, practitioners and academics working in behavioural insights – the use of behavioural and social sciences to improve decision-making using evidence-based enhancements to services, programs and policies.
Day 1
Keynote by Cass Sunstein
Issues with behavioural science ‘nudges‘ (social enticements or environmental cues to optimise desired choices, especially with respect to social policy). To force people to choose when they don’t want to make a decision impacts their dignity. This isn’t the intention of nudging.
Nudges must:
- be consistent with people’s values and interests
- be used for legitimate reasons
- not violate individual rights
- be transparent (public scrutiny incl. after randomised control trials)
- involve informed consent.
Question from floor: working in developing nations like Papau New Guinea, how can randomised control trials (RCTs) be used ethically when this violates cultural norms that some groups get a benefit and others don’t?
Sunstein: if you know something will definite have benefits, you don’t need RCT. Simply roll out the behavioural intervention.
‘The private sector narrowly allowed to have self0interested nudges (when trying to do good) but not governments.’ – Cass Sunstein
What is the repertoire of behaviorally-informed interventions that go beyond nudges? This includes sugar tax, other mandates. But which is best to choose? Identify the highest net benefits usually (such as cost benefits). Choice-preserving efforts should be to priority.
Question on ethics and Cambridge Analytica.
Sustein: It seems there was deception with Cambridge Analytica, so it wouldn’t be right to call it a nudge. Anything that pushes people into an information cocoon that adversely affects their privacy or expands a company’s own horizons is harmful.
Another question on restorative justice is not deeply discussed. See instead Professor Jessica Roberts: ‘Failing to explicitly acknowledge that some people are nudge-proof threatens to erase them from the conversation, possibly leading to even greater inequalities.’
Using Behavioural Field Experiments in Education
This panel includes Jackie Wilson as moderator, Prof John List as the key speaker and his colleague (and life partner) Dana Suskind as second speaker.
Australia is about to enter into a major educational change. A major policy focus for us is: how do we better engage parents on children’s education?
John List says that up to mid-1980s, empiricism in economics was resistant to ideas on behavioural experiments. Behaviour was seen as too complex to shift without incentives. Back then, economists acted more as ‘data sifters,’ relying on existing large-scale datasets. Education is one of the most costly and long-term investment due to hyperbolic discounting (we prefer immediete pay-offs in the present, rather than change our behaviour to benefit benefits in the future).
Note that since the 1980s, structural inequality has deepened. John List talks about Chicago where his lab is based as one example of education segregation.
Leveraging loss aversion, policy incentives have used merit-based pay for teachers to improve students’ educational outcomes. Loss aversion is the theory that people value their losses greater than potential gain. The control for these measurements is the standard bonus in the USA, given to teachers at end of school year, versus no bonus given. The intervention group were offered a cheque for $4K in September. If students don’t improve, they give money back. This has not led to large-scale changes.
List says it’s tough to implement ‘game changers’ in the education system. So instead, his team has focused on pre-school learning. In their studies, two groups of kids learn ‘executive’ and basic literacy. Another three groups focus on parental involvement, which includes a financial incentive ($7,500) or saving for kids’ university funds.
The biggest gains in 4 months was found in cognitive skills developed by preschool kids. Parent Academy works especially well. All-day KinderPrep positively impacts racial minorities. Keeping the ‘treatment group’ kids together stops education regression over the long-term.
The study is continuing to follow kids over the life course.
Dr Dana Suskind says that socioeconomic inequality starts to impact educational development from nine months after birth. Parental involvement makes a difference – it’s not about how much parents love their kids but the skills parents have available to help their kids through schooling.
The ‘3 Ts’ being piloted in early childhood education: tune in, talk more, and take turns. Addressing various behavioural barriers from day of birth, in partnership with parents and communities.
Question from the floor: what approaches work best for different ethnic or racial or socioeconomic groups?
Dr Suskind: Parent Academy has the best effect for minorities, especially for one group- Hispanics- who experience the greatest benefits (relative to Black and White groups in Chicago). They have more family networks to cover when one parent can’t make it to Parent Academy.
Prof List: ‘The welfare effect’ needs to be analysed when measuring the benefits of a study/ trial. For example, what if teachers hate what’s required to achieve, the ‘claw back effect,’ even if their students’ outcomes improve? We need to address and measure this.
Question from Sydney Morning Herald journalist: will rich White people benefit from this approach?
Prof LIst: 90% kids in the trial are below the poverty line. The rest 10% come from variable socioeconomic backgrouneds – but they also improve in Parent Academy. ‘We want every child to reach their potential.’
In the general population, high socioeconomic (SES) kids already meet their potential more often while low SES kids are being left behind. As a society, we’re missing out on their potential, their outputs. Using behavioural science to take them from not having a chance to having chance.
Second generation nudges
What are the next areas of behavioural interventions in social policy?
First up is Prof Dilip Soman. Behavioural science is more than nudges. It includes looking at the choice architecture, design, market for self control, and better decision-making.
How do we build behaviorally informed organisations? Basic example issue is, how do we get people to read social media privacy policies?
What is our role as behavioural scientists? Going from problem solver ➡ to auditor ➡ to designer ➡ to chief strategist.
Now Kate Phillips. Most behavioural science insights in social policy have been applied in implementation and evaluation. For example, behaviourally informed text messages for vaccinations and hospital reminders and fact sheets for consumers.
We now need to move into using ehavioural scbience to policy creation. But how? Randomised control trials can’t be used to generate policy. Use counterfactuals – apply BI principles to policy design. For example, changes to cigarette packages.
Behavioural science policy design was used in Victoria to improve information sharing between front line workers in domestic family violence. They value empowering survivors. There were physical barriers which limit information sharing. For example, sitting behind registrar at court.
Now Liam Smith- a psychologist who says he will pretend to be a sociologist for his talk!
Systems change theory is useful. We need to consider three things to sustain change:
- Engaged individuals
- Political opportunity
- Organisational infrastructure
Sociology theory tells us that most traditional behavioural science nudges hurt structure and agency. But these can be used to improve ‘second generation’ tools to better impact moral and social licence to take action. For example, where attitudes are weak – such as in workplace safety.
Where attitudes are strong, we need to instead focus on behavioural changes that are simple but effective. Reinforce identity and self efficacy to shift attitudes. For example, in water saving behaviour.
Now Elizabeth Hardy from Canada. Their first generation of behavioural science policy work was on traditional touch points (emails or phone calls) that increased public participation in surveys and charitable giving. Then behavioural insights worked on educational savings incentive trials.
Second generation nudges is what is now being tested. In March, the Canadian team trialled adding education savings to existing birth bundle, adding only 20 minute to the online sign-up process.
Next, Canadian BI teams will move to trial letters from principals to parents, as well as offering via social assistance.
Now the speakers come together in a panel and audience question and answer.
Soman: We need to move away from talking about ‘failed’ trials. Whether the intervention worked as we intended or not, we still learned something valuable. Nudges can be thought of in 2 ways.
- Systemic changes (simple nudges).
- Different sub-groups. Need different nudges or approaches. E.g. superannuation calculator to help individuals weigh up individual choices to make decision on their preferred plan
Question from the floor: what are some future areas to apply second generation behavioural science tools?
- Social cohesion
- Preventative health.
Fire Side Chat with Cass Sunstein
Next session is the ‘fire side chat’ between Cass Sunstein and Tara Oliver, Managing Director of BETA, the Commonwealth behavioural insights team based in Canberra.
How ‘Nudge’ was born: thorugh a fairytale. In early 1980s Cass Sunstein was an English major battling ‘monsters chanting “People are rational.”‘ Best description of classical economics!
Prof Sunstein manages to make an analogy between behavioural science insights and Star Wars!
No matter what nudges we receive, either from Sith or Jedi, people have a choice. Another theme from the films that relate to BI is that a parent would do anything to save their children.
Nudging the nudgers: how do we stop policy makers from falling prey to their biases (i.e. choice architecture for choice architects)?
First, public consultation ‘is not Kabuki theatre.’ It matters. People read comments others make. Can’t ignore them.
Other tips on nudging the nudgers:
- Invest in cost benefit analyses
- Maintain scientific integrity
- Commit to retrospective analysis. Learn after fact how any intervention is going.
Question: Technology can be used for nudges in ways that may ‘cost’ government. For example, apps that help you avoid parking fines means less revenue.
Prof Sunstein: Cost benefit analysis should be for social good and public interest. Fines shouldn’t be about revenue. It’s to help us comply with law. Policymakers need to weigh up social costs versus social benefits for the public good.
Question: Is Trump administration still committed to behavioural science as was the Obama administration?
Prof Sunstein: Yes. For example, they’re interested in calorie labels, food safety and nutrition work.
Questoin: Are subsidies and mandates still being used in current climate?
Prof Sunstein: Yes government agencies continue this work. Manipulation by deception of consumers is outlawed by many countries. Some governments still use manipulation edging towards deception, including for health nudges. We need to think further on this.
How can we follow Prof Sunstein’s first point on bill of rights for nudging when everyone’s values and interests can be different?
Prof Sunstein: if it’s an ‘easy’ issue, by using personalised preferences. If it’s complicated issue do cost/ benefit and address people’s losses. For example, use reminders or easy opt outs for people to minimise their losses.
Day 2
Scalability and economics
Prof John List on the ‘scalability problem’ (results from trials ‘diminish substantially’ when rolled out to broader population): post-study probability issues; publication bias; trial participants more compliant; other orgs less methodical than researchers.
Academics want to publish and independence. But financial markets and industry want commercial outcomes. We need leaders willing to admit they don’t have all the answers and who’ll commit to continuous testing. Yet the reality is that testing & re-testing is expensive.
Question from floor by Price Waterhouse Coopers starts with, ‘Everything we do on cyber security is failing.’
Early work with Harvard Uni testing loss aversion (what can you lose if you click on this link) and shared responsibility (your colleagues could be hurt if you click on this link).
People who have poor financial well-being but don’t understand this are of highest concern to behavioural insights practitioners. Financial well-being measured by things like having pay day loan or showing pay lower than expenses at end of the week.
Nudging the nudgers
How do we overcome our cognitive biases as researchers and practitioners?
Varun Gauri, behavioural economist from the World Bank is now talking. He argues we should give policy makers an opportunity to think through their biases. The World Bank surveyed staff showing their biases. Then they reproduced the study in the UK. The World Bank and other civil servants were surveyed. Their biases:
- confirmation bias;
- gain/loss framing;
- sunk cost bias;
- risk taking for others;
- mental models of poverty.
Confirmation bias: Political views and expert training of pubic servants leads to errors. Financial incentives backfire, as public servants can have a strong identity, moral conviction, and high education/ scepticism that can make them resistant to evaluating conflicting information.
Government officials make the same errors we see in other behavioural science studies with other populations, including on taking risks and in misunderstanding poverty even when they’re experts on poverty.
How to address cognitive biases of public servants and other officials? Approaches are similar to other behavioural science interventions, including deliberation.
The World Bank is undertaking research on de-biasing gender equity in hiring and knowledge management.
Follow results of this work on Twitter, using the hashtag #embed_wb.
David Halpern CEO of the Behavioural Insights Team discusses forthcoming work on UK policy makers’ biases. They don’t always allocate attention according to need. They are susceptible to drawing conclusions based on how policy is framed (how the wording of a social problem is presented).
Biases of policy makers includes overestimating how people like them will act (according to their individual expectations). Senior leaders are especially susceptible to over confidence.
An OECD study shows 57% of economists dislike interdisciplinary collaboration. But collaboration can guard against biases!
How to overcome behavioural biases in public administration using behavioural science? This includes assembling cognitive and conceptually diverse teams.
Other cognitive biases in public administration include motivated reasoning due to expertise; present bias; and preference for external consultation without rigour.
Peer review and pre-mortems might help. We need to build institutions outside of government to mitigate public servant bias.
Incredibly important question from the floor about how de-biased methods of recruitment which currently focus on gender equity* can be extended to disability. (*I note these are problematically conceived as a gender binary.)
The answer is: essentially the work hasn’t been done yet.
Violence and crime: putting behavioural science on trial
Assistant Professor Anuj K. Shah begins this next panel.
We tend to study criminal patterns but not the decision making processes behind these patterns. We should be looking at creating moments of choice. For example, 41% people in New York City don’t show up for court. This means that low level offences (such as fines) become bigger.
When we know the behaviour and the window of time when we need to intervene, we can use behavioural science “nudges,” such as by redesigning summons and text reminders.
Other behaviours tougher to identify because we can’t be there to preempt decisions being weighed up before a crime.
Other more complex behavioural science interventions being tested are to teach youth who are engaged in crime to question their assumptions about their behaviour.
The illusion of transparency is also being tested to reduce anonymity as a crime deterrent.
‘We must look into how moments of choice arise or don’t.’
Mary Ann O’Loughlin on Behavioural Insights Unit’s work on to address reoffending on domestic violence offences.
For example, simplifying the language of Apprehended Domestic Violence Orders (ADVOs) and correcting social norms on gender violence.
Other behavioural science trials on domestic and family violence by Behavioural Insights Unit:
- Court Attendance SMS
- What’s Your Plan co-designed with Aboriginal experts.
- AVOW Digital App to help perpetrators understand ADVOs.
Read more about the first Behavioural Insights Unit trial on reducing domestic and family violence reoffending on the annual report.
David Yokum talks next about his team’s randomised control trial on 1,000 police officers wearing body cams in District of Columbia. The results are unexpected. The use of force varies between police wearing cams and those who do not, but only to a small amount, and the diference is not statistically insignificant. Body cam police receive slightly more complaints about their conduct and they are more likely to be prosecuted – but this is a null result (not statistically significant).
So are body cams worth it? Yokum says the null results can be interpreted in different ways. For one, even though it’s not statistically significant, it might make a difference in one or two cases. This can have a big community impact. The study’s outcome have been enough to shift policy discussions.
Yokum’s team, The Lab @ DC, invested heavily in engaging communities in their science. They’ve been sharing ongoing results and explaining technical aspects of scientific process. Strong public outreach via local media and their website.
One of my favourite behavioural science practitioners, and project lead on the What’s Your Plan trial by the Behavioural Insights Unit, Eva Koromilas (pictured below) was asked by the panel to describe how the trial worked with Aboriginal client and community support officers and domestic and family violence reoffenders.
Why do other areas of state government not have their own behavioural science teams? Mary Ann O’Loughlin says bringing together a team of multidisciplinary specialists takes time and it’s also useful to have them in a central place collaborating with other areas.
New Frontiers in behavioural insights
Around the world, behavioural science is having major impact, from improving organ donation in Canada, to the new health policy to address childhood obesity in UK, to the sugar tax introduced in Australia yesterday. These are all outcomes of behavioural science.
David Halpern who set up the first behavioural insights team in the world (in the UK) discusses how the Behavioural Insights Unit (Australia) trial, What’s Your Plan?, as an example of innovative use of behavioural science for policy and service delivery change.
The Other Sociologist 