An artificial conscience shaped by human values
VirtualMorals is an academic study and machine learning project aimed at analyzing the trolley problem and its variations, modeling public opinion for each variation, and understanding why responses differ between them.
Help influence the model!
What this is
We plan to accomplish the following goals with our survey and analysis:
With each participant’s responses to a subset of questions randomly chosen from the entire list, we would like to perform the following tasks:
- Develop a model that uses the estimated public opinion to determine a solution for moral dilemmas.
- Examine trends among questions with similar responses.
- Mathematically group participants into different ethical schools of thought regarding “trolley problem” questions.
- Discern what people value most in ethical dilemmas. Which values and principles are most important to us, and which ones are we willing to sacrifice?
How it works
Our plan includes these three steps:
- Survey the public on different variations of the trolley problem.
- Develop a model to group participants by ethical schools of thought.
- Extrapolate data to discover which factors influence public opinion the most.
Why it matters
Our reasons for creating the model are twofold. Namely, we hope to:
- Extrapolate useful insights from ethical surveys when it is infeasible to ask each participant every question.
- Take the first steps in creating artificial moral agents (such as autonomous trains) that act on public opinion when faced with difficult decisions by creating a machine learning model to predict public opinion.