We know that AI has been trained on human input. But does it work the other way around as well? Can humans be trained to adopt AI points of view? A paper published in October last year by Lucía Vicente & Helena Matute might have identified evidence of human inheritance of AI bias.
Here is the gist of their research:
- The research paper is titled „Humans inherit artificial intelligence biases“.
- The authors conducted three experiments to investigate whether people who perform a medical diagnostic task assisted by a biased AI system will reproduce the model’s bias in their own decisions, even when they move to a context without AI support.
- The results showed that the biased recommendations by the AI influenced participants’ decisions, and when those participants, assisted by the AI, moved on to perform the task without assistance, they made the same errors as the AI had made during the previous phase.
- Thus, participants’ responses mimicked AI bias even when the AI was no longer making suggestions.
- These results provide evidence of human inheritance of AI bias.
It sounds „shocking“ at first. But then again, let’s dive into it a little and see where the study may fall rather short or may be improved:
- For example, the study only looked at one specific type of task (medical diagnosis) and one specific type of AI bias. It is possible that the results would be different for other types of tasks or other types of AI bias.
- Additionally, the study only looked at a small number of participants, so it is unclear whether the results would generalize to a larger population.
- The study did not investigate ways to mitigate the effects of AI bias on human decision-making. Future research could explore ways to reduce the impact of AI bias on human decision-making, such as providing feedback to users about the potential biases in the AI system.
For me personally, the whole point this study drives home is that human input, expertise, and critical thinking will be a winning asset in our common AI-future. More than ever will we be challenged to really apply our well-defined thinking to what AI brings to the table and NOT just consume its output without second thoughts.
Let’s always keep a healthy distance to Bill.