The veil of ignorance experience - a proposal

[Disclaimer - there are hypotheses in this proposal and a lot of areas where I'm sure existing research exists. I'm publishing now as I believe the idea itself is strong enough and as I would benefit from anybody sufficiently interested to contribute to the effort, including to conduct the necessary scan of existing relevant research.]

Summary

This project is to design an 'ideal' future, and one of the main 'ideals' is fairness. I believe that Rawls' Veil of Ignorance thought experiment is one of the most powerful tools available to simultaneously inform what we believe a 'fair' future to look like, and to shift attitudes and behaviours towards equality and redistribution in the present. I propose developing a mass participation experiment, a la MIT's Moral Machine, to bring this thought experiment to a general audience and develop a large dataset that will provide a shared vision for a fair future, among many other insights. I hypothesise that the average Gini coefficient - a simple of measure of distribution/inequality - for the fairest global income distribution will be <0.2 (it was estimated at 0.61-0.67 for 2005), and that participants will commit to backing more redistributive policies having participated in the experiment. It will take a significant investment to do this idea justice - if you are interested in funding this work, building the web application, or driving the research, please contact me.

The purpose of What Future World? is to envision and design an ideal future. There are many aspects of an 'ideal' future that people would debate, but one thing we can say is that an ideal future would be fair and just. Those terms mean different things to different people - as we will see - but we can agree the concepts represent 'ideals'. So how do we begin designing a 'fair' and 'just' future when we know people have different visions of what those terms mean? It turns out that a philosopher by the name of John Rawls came up with a thought experiment over fifty years ago to help us do exactly this. It's called, the veil of ignorance.

What is the veil of ignorance?

In Ethical School Leadership, Spencer J. Maxcy writes:

Imagine that you have set for yourself the task of developing a totally new social contract for today’s society. How could you do so fairly? Although you could never actually eliminate all of your personal biases and prejudices, you would need to take steps at least to minimize them. Rawls suggests that you imagine yourself in an original position behind a veil of ignorance. Behind this veil, you know nothing of yourself and your natural abilities, or your position in society. You know nothing of your sex, race, nationality, or individual tastes. Behind such a veil of ignorance all individuals are simply specified as rational, free, and morally equal beings. You do know that in the “real world,” however, there will be a wide variety in the natural distribution of natural assets and abilities, and that there will be differences of sex, race, and culture that will distinguish groups of people from each other.

Our tendency as rational, self-interesting human beings makes it difficult to be truly altruistic. Even if we believe that greater economic equality is 'fair', it is very difficult to make actions fully congruent with that belief. Implicit in many government policies is a belief of what a 'fair' distribution of income and wealth across a society should be - both domestic and international.  For a simplified example, a US bill to reduce foreign aid is implicitly advocating for a high level of global income inequality. They believe that is fair.

Herein lies the problem that Rawls identified. Humans are seemingly terrible at removing our current position in the world, in terms of income, wealth, status, and power, from our perspectives over what is fair. When making claims like "it's fair that I earn 10x more, I worked very hard for this" we forget two vital factors: 1) people that can say this have often won the 'lottery of birth' and can claim no responsibility for its benefits, and 2) social mobility is significantly less common than is often cited.

The lottery of birth

By nature of not existing beforehand, we have no say in the conditions of our birth. We have no responsibility for our gender, race, health, or a number of personal attributes driven by nature over nurture. We don't have a say in when or where we are born, or who we are born to, meaning most of the nurture component of our early identity is also down to luck. It is obvious that the majority of the attribution for our lot in life comes down to these factors - someone born as a woman in poverty in Afghanistan won't have the same opportunity as someone born a white male in New York City. It is inherently unfair that we have inequality based on luck. When we design a 'fair' future, Rawls says we must imagine we play the lottery with all of these attributes and be happy with the society we have designed no matter who we turn out to be.

The myth(?) of social mobility

Many would say the American Dream - the notion that anyone can go from poverty to millionaire through hustle alone - seems appropriately branded. A dream, not a reality. Only 43% of US adults surveyed said that there is such a thing as the American Dream. For those under thirty, the figure falls to 29%. One definition - children being economically better off than their parents - has been in decline in the US since the 1940s driven primarily by increasing income inequality over that time. While I can't find great studies on social mobility, there is great research showing that direct cash distribution to those in poverty in the US lead to improved health and education outcomes, and reduced crime, without a fall in hours worked. Strong social mobility would suggest that poverty is a choice - that with enough hard work anyone could pull themselves into a comfortable life. I believe the weight of evidence is against this view. There must be structural barriers preventing people climbing the income distribution because when these are taken away - through progressive policies - these individuals are very capable of improving their situation. This is to say nothing about social mobility on a global level.

I plan to update this article as more research is published on social mobility at a macro level, and hopefully this proposal can also contribute to the research agenda by determining exactly what data is needed to test the most significant hypotheses. I'm publishing this proposal now as I have conviction in the hypothesis based on the less-than-ideal evidence available, and that the final conclusion and recommendation weigh little based on this argument. 80% of people in the US are better off than two-thirds of the global population - it is almost a truism that anyone at the top of the global income distribution will be shocked by the lottery of birth alone.

We see then that a lot of the attribution for an individual's lot in life is based on luck, and yet it is difficult for those in the upper end of the income distribution to accept as much. In order to design a truly fair future, we need to strip as much of this cognitive bias as possible before someone begins their design. We need people to internalise the belief that 'I could have been born anybody in the world', for all that means, and design a fair future world accordingly. This is the goal of 'the veil of ignorance experience'.

The veil of ignorance experience

Our goal in designing a 'fair' future is two-fold: develop a shared definition of 'fair', and shift individual attitudes towards equality to help us move towards this fairer future, for example through persuading more people to vote for progressive, redistributive policies. MIT's Moral Machine has shown us that mass participation online thought experiments can both inform a large audience and develop a democratic approach to answering a complex problem, suggesting this form of mass-participation thought-experiment could prove successful for us too. Here I'll outline the overall experience before providing the details of the most important steps.

Flow of the experience

• Introduce the experience and define all key terms
• Ask them to design their ideal income and wealth distributions, and answer some survey questions
• Ask for their demographic information
• Share their current state diagnostic
• Ask them if they want to redo their ideal distributions
• Rerun the diagnostic on their distributions
• Ask them for further refinements.
• Iterate until participant is happy with final distribution
• Provide information on what it would take to achieve, what they can do to help make it a reality

Testing/retesting

Our second goal is to observe a change in attitude for having participated in this experiment. The best way to prove this is through testing the participant's attitude and behaviour towards 'fairness' before and after the experience. We accomplish this by asking for their initial, 'fair' distributions before anything else, and comparing it with the final distributions they submit. We can also ask them survey questions about their attitude towards redistributive policies. This is better if we can ask about actions 'how did they vote on X' rather than pure attitudes. We could follow this up with longitudinal surveying asking how they plan to and then how they did vote on certain policies, or how much they personally donated compared to previous years, to see if the attitude change stuck

Designing ideal distributions

Economists typically measure fairness using two distributions - income and wealth. We measure the income level at points across the entire spectrum, from the lowest to the highest, to see how evenly is the measure distributed. The most 'equal' society would be one where everyone has the same income/wealth, the most unequal where only one person has it all. This is not to say that most equal is the most fair - many believe that a certain level of inequality is justified and fair. With this experiment we are trying to identify what that level is. For the sake of simplicity, I would propose using the Gini Coefficient to measure these distributions, and a simplified version at that. The beauty of the Gini coefficient is that it provides a single score on which to judge inequality. A score of 1 shows perfect inequality, and 0 shows perfect equality. For context, the distribution of global income was estimated at 0.61-0.67 for 2005, and the most 'equal' societies tend to be in Eastern Europe and score ~0.25. There are problems that come with this oversimplification, but I don't believe any of them are material for this exercise.

This score is typically calculated using a mathematical formula that accounts for every data point in a population, i.e., the wealth or income of every person in a country. For further simplification, I propose breaking society down into four groups for our experiment, a classification proposed by Thomas Piketty in "Capital". The bottom 50% of a population called the "lower" or "working class", the middle 40% ("Middle class"), and the top 10% split into the next 9% ("well-to-do class") and the top 1% ("dominant class"). It would be too much - cognitively and operationally - to ask people to design a full income distribution. It is reasonable to ask them how much income should be earned by each of these four groups and ensuring it sums to 100%. It is simple enough to say "I think the lower 50% should earn 30% of income, the middle 40% should earn 45%, the next 9% should earn 18%, and the top 1% should earn 7%" and be told that corresponds to a Gini coefficient of ~0.3. To me, this seems the optimal point in the trade-off between accuracy and simplicity.

Current state diagnostic

Once we've captured their initial, 'uninformed' design for the purposes of proving the change we will create, we ask for their demographic data and share a 'current state diagnostic' with the participant. The first step would look very similar to the calculator produced by the World Inequality Database. Using their demographic data, it tells them where they fall on the income or wealth distribution of their country, and for the world. This will likely be eye-opening for most participants but is insufficient to create change. It is too easy to observe where you are in the distribution, often be surprised at how low the income or wealth is for the vast majority of the world and then...go back to your normal life with a couple of interesting facts to bring up at the dinner table. This is where we need to leverage the power of the lottery of birth.

Once they've learned all they can about their current position on the distributions, we share the concept of the 'lottery of birth', and ask the participant to 'play the lottery'...and find out who they would be if their luck had been different. What if you were born a black woman in Burundi and not whoever you actually are? Here is where we show participants what that would really mean. We would use population statistics to determine the likelihood of being any one person. We need to determine how many demographic attributes are necessary to paint an accurate picture - what things matter? Our World in Data has a superb article showing that what matters most for your living conditions is not who you are, but where you are, meaning we would likely be okay with just country of birth, gender, and household income/wealth at birth as my hypothesis for the three most important pieces of data. For example, in the present day, there's roughly an 8% chance you will be born a woman in India with average (for India) family wealth. It adds complexity, but I think the experience would be more effective to present the hypothetical life as the same age as the participant. They can then make direct comparisons about their living conditions without mental gymnastics. To do this we would just need to use statistics from the participants birth year rather than the present day.

This alone would be extremely powerful. To take it a step further would be to further communicate the 'expected life' of that individual. The existing approach would say that in the present the hypothetical individual is most likely in the X'th percentile on the global income and wealth distributions. We could paint a fuller picture, if the data is available, to look at the probability distribution associated with that mean position. The OWID article does this brilliantly. The mean position on income distribution for the average person in Ethiopia is significantly below that for the average person in Denmark, but even more interesting is that even the very top of the income distribution for Ethiopia is the lowest point of the distribution for Denmark. We could say for example that even if you were to be the top 1% of black women in Burundi - whether you attribute that to hard work or sheer luck - you will still be worse off than the vast vast majority of people in France, for example. The full distribution would be too much information, so we'd likely pick the most insightful descriptive statistics, like the example just given.

We can then also show how their initially designed income and wealth distributions stack up against the world as it is today, and the diagnostic can be run against either the real world or their initial design. The more we can get the participant to empathise with this hypothetical person - the person they could have been if they had different luck - in either the real world or their initial perception of a 'fair' world, the more likely they are to want to improve outcomes for them.

After this experience, I believe the instinct would be to 'play the lottery' - rerun the diagnosis - multiple times. Targeting this experience at people in the global 9th decile would mean that 90% of the time a story will be painted of a life much worse off than their own. It will be incredibly rare for someone to return a hypothetical individual significantly better off than themselves. This needs to be thoughtfully designed to build up the weight of evidence on each re-roll to make it unambiguously clear how lucky they have been in the lottery of birth. I believe it will be a combination of empathising with individual 'alternative' lives and an overwhelming weight of evidence from multiple rolls of the dice that will cause a change in participant attitudes.

Refinement

After this critical learning moment, we give the participant the opportunity to refine their 'fair' distributions, expecting that the average participant would design something significantly more equal. From here, we enter an iterative cycle of testing the design through the lottery diagnostic until the participant is happy with their 'fair' design and submits it to the database of 'fair' income and wealth distributions. It is this database that will be used to achieve our first goal of creating a shared definition of 'fair'.

Providing information for action

Two central principles of What Future World? are pragmatism and optimism. We don't want participants thinking this was an inconsequential exercise. We want to provide them hope that this ideal, 'fair' future can be obtained by giving them practical actions they can perform in the period immediately after the experience. This will reinforce the lessons from the experience at an individual level, and will lead to societal change through greater civic participation. It will be difficult to provide specific actions for people in different circumstances all around the world, but it is important enough to be a thoughtful component of the experience.

Hypotheses

In short I hypothesise that the experiment will be a success! In particular I hypothesise that the average Gini coefficient for the fairest global income distribution will be <0.2, more equal than any country in the world today and starkly more equal than the global income distribution currently estimated >0.6. I believe that average Gini coefficients will drop significantly between initial design and final submission, and that participants will commit to backing more redistributive policies having participated in the experiment.

A questions this raises

How likely someone is to be at their position on the income/wealth distribution based on their demographic factors? This looks at the question in reverse. What demographic data could we use to make comments on the social mobility someone has experienced, or on social mobility as a whole? Use me as an example. I won the lottery of birth in many ways: I was born white, male, and healthy in England in 1993. A very rough estimate says I had a 0.2% chance of being born with that profile. Or a 1 in 500 chance. It's likely significantly longer odds that I'd have above average intelligence, athleticism, and mental/physical health disposition. That said, we didn't have much growing up. My household was in the bottom 20% of the income distribution when I was born - a proxy for my 'starting point' in life. I appreciate how much I have won both the lottery of birth and of social inequality - and I think the data would give me, and plenty others like me, an even greater appreciation.

We would then need to be incredibly careful about how we communicate this information, particularly those doing much better and much worse than predicted. While wanting to communicate that the variance is still almost entirely driven by luck, there is a fine balance between saying "It's all luck, so what's the point of being good and/or working hard" and "Being good and working hard are all that matters" - a la the American Dream. This is the crucial philosophical question we hope to narrow in on, but would never suggest we could solve completely.

Risks

I think it is almost inevitable that, if successful, this project would evoke a very strong backlash from a lot of people who won the lottery of birth. Successful people rarely like to be told that they can't take the credit for their outsized success, and some will inevitably take it much worse than others. This in and of itself is not a risk, rather it is a measure of success. Such a backlash would suggest this idea has gained sufficient awareness and traction to be threatening existing power structures, structures that would inevitably need to change between now and our ideal, 'fair', future world. The risk comes from whether these powerful incumbents are able to leverage their resources to quash or restrain progress to a fair future. In my opinion, one of the most difficult problems facing our progress is how to persuade people with power that the world we want requires us to sacrifice power. I think that this project would be one of the, if not the, best way to achieve that aim. It is therefore worthwhile to take this risk, execute the project in a thoughtful way, and mitigate the risks before they emerge.

To action

Hopefully you think this is a good idea - if not, please leave a comment or complete the feedback form. How we can make this a reality? I am happy and excited to lead this project under the What Future World? banner - there is a very clear alignment in purpose. There are a few areas I need support to get this off of the ground. The first is the money to build and maintain the platform, and fund research. If you're in a position to financially back this project please reach out - or expect me to reach out in due course. We would then need the research capabilities - including big data analytics - to determine what is possible and build the academic rigour behind the experiment, the design capabilities to create a thoughtful, engaging, and effective experience, and the technical capabilities to make it all a reality. If you can contribute any of those things, or think you can offer something else of high impact, please reach out.

I truly believe Rawls' idea, with a sufficient boost, is powerful enough to change the world.

Please share your thoughts if you have any feedback on this article, or leave a comment below.