At our All-to-All meeting on April 24, 2024, Senior Scientist and Head of ATAP’s Fusion Science & Ion Beam Technology, Thomas Schenkel, presented a fascinating and insightful talk on the risks of human cognitive biases being reproduced or even amplified in Artificial Intelligence (AI). Cognitive biases refer to attitudes, prejudices, or judgments we unconsciously hold about people or groups.

Schenkel noted that there are four cognitive biases in AI as well as machine learning (ML), a subset of AI:

  • Selection Bias – selecting data not representative of the entire population
  • Reporting Bias – people’s tendency to underreport information
  • Implicit Bias – people’s unconscious tendencies to assume
  • Framing Bias – the tendency to be affected by how information is presented

He then provided five best practices to enhance objectivity in AI/ML systems:

  • Always select a random sample (instead of the first hundred data points)
  • Verify your data sources through comparison with other data sources
  • Assign more (diverse) scientists/engineers to develop the AI/ML system
  • Establish proper peer review procedures to cross-examine logic and unconscious bias
  • Prioritize objective and factual data over subjective (usually qualitative) data

The presentation also referenced a study that evaluated the emergence of ethical guidelines for AI. The study entitled “Diversity, Equity, and Inclusion in Artificial Intelligence: An Evaluation of Guidelines” evaluated 46 ethical guidelines for AI published from 2015 to 2022, fleshing out 14 diversity, equity, and inclusion (DEI) principles and the 18 DEI practices recommended underlying these 46 guidelines.

While the authors say the guidelines often include DEI principles and practices, they found that “little is known about the DEI content of these guidelines, and to what extent they meet the most relevant accumulated knowledge from DEI literature.”

In conclusion, they recommended that practices for implementing DEI principles in AI “should include actions aimed at directly influencing AI actors’ behaviors and awareness of DEI risks, rather than just stating intentions and programs.”

Click here to learn more about cognitive biases in AI/ML.



For more information on ATAP News articles, contact