Survival of the Best Fit
Bias & discrimination
Hands-on experimentation
What can I learn?
A 5–10 minute browser game where you play a startup CEO automating hiring. The algorithm visibly inherits bias from training data, producing discriminatory outcomes despite neutral-seeming design choices.
Core insight
Bias does not have to be intentional to be harmful — an algorithm trained on biased historical data will reproduce and amplify that bias automatically, even when the designers had good intentions.
How to use it in daily work
Play it yourself first, then use it as a conversation-starter or demonstration tool in sessions on AI and discrimination.
- After playing, you can explain to a client why an AI recruitment tool might systematically disadvantage them — with concrete cause-and-effect understanding, not just a vague claim.
- Works well as a 10-minute opener for a group session on AI and jobs.
Note
The bias demonstration is scripted rather than a live ML simulation — powerful pedagogically but not a technically accurate model.