Episode 11 - AI Has Biases, So Do We: Judgment in an Age of AI
AI is getting remarkably good at producing answers, but producing an answer and exercising judgment are not the same thing. In this episode of “Decoding Ethics,” host Jeanne Viljoen speaks with Sir Andrew Likierman, Professor of Management
Practice in Accounting and former Dean of London Business School, about what that distinction means for professionals working alongside AI.
Andrew, author of “Judgment at Work: Making Better Choices”, in which he sets a six-element framework for understanding and improving judgments, explains why he still holds the view he set out in his 2024 Forbes article “Good News for
Human Beings – AI Doesn’t Do Judgment.” He identifies awareness and trust, along with feelings and beliefs, as the human elements of judgment that will matter more as AI advances.
Because we don’t know what has gone into AI’s “wonderfully fluent” answers, Andrew argues that professionals must learn new skills, including asking the machine to state its sources or methodology. AI can challenge our biases,
but it has its own: the main AI agents he asked all said they are more positive than negative because they have been programmed that way. When AI and the professional’s experience point in different directions, he says to treat AI output as
you would a colleague’s opinion: “just one more source of potential information.”
The conversation also covers the role of values in professional judgment, why the traditional apprenticeship model needs rethinking and Andrew’s four recommendations for using AI to strengthen, rather than weaken, judgment.
Takeaways
- Judgment is more than decision-making. Sir Andrew Likierman defines it as combining personal qualities with relevant knowledge and experience to form an opinion or take a decision.
- The human elements of judgment will matter more in the AI era. Awareness and trust, along with feelings and beliefs, are the particularly human elements, and even knowledge depends on the context that humans supply.
- AI output has to earn trust. We don’t know the programming, training data or quality control behind it, so professionals need to ask supplementary questions and request sources or methodology.
- The goal is awareness of bias, not freedom from it. AI can challenge our biases, but it has biases of its own, including being programmed to be more positive than negative.
- Treat AI like a colleague’s opinion. It is one more source of potential information, subject to the same professional skepticism.
- Judgment must be developed more deliberately. With AI doing more of the foundational work, the traditional apprenticeship model needs rethinking.
Watch the podcast here.




.png?sfvrsn=dd675fd5_2)

6a6c7fe2013f4be3b0ecc0a5930da5c9.jpg?sfvrsn=ab5334c8_0)

/business-management-global-connection/istock-1167579720-c.jpg?sfvrsn=ff93f9a5_2)
/audit-assurance/istock-1169206203-c.jpg?sfvrsn=1d6f9b25_6)


/ethics-and-professionalism/istock-1141115724-c.jpg?sfvrsn=4e54d691_2)

/audit-assurance/istock-818732836-c-v3.jpg?sfvrsn=ae44e7b7_0)

.png?sfvrsn=905ee1bd_0)
/legal-secretarial/istock-866706340-c.jpg?sfvrsn=d7f57b8c_2)