The Moral Machine

An experiment for crowdsourcing assessments of morally significant outcomes in hypothetical autonomous vehicle ethical dilemma scenarios. We collected over 40 million decisions from around the world at time of study. We added a "Classic" section too, where visitors could be surveyed on human-agent ethical dilemmas.

I contributed significantly to the design process, and in development of many features, including the design wizard, scenario browser, result screen, share panel, cloud migration, and language internationalization.

We also studied assignment of blame in the case of override-capable human+AI piloting situations using a limited-pool study, and crowdsourcing of ethical decision-making.


HTML5/CSS3/Bootstrap + MeteorJS/NodeJS/BlazeJS + MongoDB + AWS S3 + Meteor Galaxy + i18n + D3.js


MIT Media Lab, 2016-2017



MIT Master of Science in Media Arts and Sciences thesis: Dsouza, S. (2021) Crowdsourcing moral psychology.



Try out the Moral Machine, the Moral Machine Classic, and the interactive dynamic country preferences visualization explorer now, at, or try out only the Moral Machine country preferences visualization explorer at my repo backup here.

Explore relevant MIT Media Lab project pages on the Moral Machine here, on human+AI blame assignment studies here, on ethics of autonomous vehicles here, and on crowdsourcing of ethical decision-making here.

Find media coverage:

A behind-the-scenes look at the Moral Machine project can be found at Nature Social Science.

Scenario assessment screen

Classic trolley problem screen (in Arabic)

User-generated scenario design screen

Mockup of human+AI override-capable vehicle blame assignment study stimulus

Exhibited at the MIT Museum

Exhibited aboard the MS Wissenschaft