The ability for autonomous vehicles to make morally acceptable decisions is something that has troubled technologists and ethicists alike in recent years.  The trolley problem has exemplified the challenge we face in developing a technology that is digestible to the wider public. Perhaps the biggest exploration of the issue has recently born fruit with the publication of a study documenting the outcome of 40 million decisions made by millions of individuals around the world.

The work, which was led by MIT but also contained support from Harvard, the University of British Columbia and Université Toulouse Capitole, saw volunteers participating in ethical tasks hosted on the Moral Machine website over an 18 month period.

Moral decisions

The experiment was able to identify three core moral criteria people use when making such judgements: human lives are worth saving more than animal lives; it’s important to try and save as many lives as possible; and younger lives should be saved over older ones.

A number of less acceptable trends emerged however.  For instance, larger people were around 20% more likely to be killed than fitter people, whilst poorer people were 40% more likely to be killed than rich people.  There were also some interesting cultural differences emerging. For instance, in countries with strict laws and institutions, people were much less likely to save pedestrians who were crossing outside of designated areas.

The data, which is publicly accessible, will hopefully guide not only the industry as they develop autonomous technology, but also governments when they craft the policies and regulations surrounding autonomous vehicles.  Despite this potential, and the huge depth of contributions from countries around the world, the researchers do accept that people volunteered, and therefore they cannot guarantee that the data is representative. Nonetheless, the hope is that the next phase of autonomous development will benefit significantly from the data.