Good Leaders Can't Be Biased, Right?

As leaders, we are accustomed to making many decisions daily. Yet, how often do we assess the nature of those decisions and the unconscious biases that may influence them? Good leaders can’t be biased, can they? 

Daniel Kahneman and Amos Tversky first introduced the idea of cognitive biases and their impact on decision-making in 1974. Today, more than 180 different forms of human biases have now been identified. 

Everyone develops biases during their lives. They help us quickly process information. But behavioral scientists have illustrated that our decision-making processes are often swayed by preconceptions that come from our culture, experiences and the environment we grew up in.  

There are many different kinds of ‘unconscious’ bias that exist within workplaces. We exhibit “affinity bias” when we seek those with similar thoughts and behaviors, either within our team members or with other leaders, and try to relate to them. “Halo effect” occurs when we like a person, company or viewpoint and we’re biased to think everything about it is good. While making talent decisions, we can stereotype the person based on their gender, race, height, clothes, voice and behaviors. This is “perception bias.” And when we think certain members of groups are likely to succeed or fail, we exhibit “confirmation bias.”

When these unchecked biases cross over into AI technology development, they can create bias in the training data, discrimination in labeling as well as under sampling or oversampling of the data set. For example, consider a university admissions process that uses AI decision-making in its application procedure. The data used to feed the AI model could consist of admissions data going back 40 years. Except 40 years ago, admissions were heavily weighted against women and immigrants. The AI application is not introducing any new bias, but it is subject to the inherent bias that exists in the training data. Biases in training data, due to either prejudice in labels or under-/over-sampling, yields technology models with unwanted bias. Leaders and application developers have to be cognizant of such biases in order to develop models without bias. 

If diversity of thought is essential for innovative, engaged, inclusive workplaces, all leaders have to persistently work to better understand and eliminate bias.

Here’s a few ideas for leaders to start evaluating bias in decision-making:

Evaluate and challenge your own level of bias: Not all first impressions are correct. Harvard has developed a quick implicit association test for people to be more aware of their unconscious biases. Leaders can use this test to understand where their biases come from, then talk about unconscious bias with others and challenge their own assumptions. 

Machine learning model developers and team leaders can evaluate bias through IBM’s AI Fairness 360 Toolkit which provides bias mitigation algorithms, fairness metrics, demos, videos and a tutorial. As our CEO, Arvind Krishna, has outlined, we believe all vendors and users of Al systems have a shared responsibility to ensure that Al is tested for bias, particularity when used in law enforcement, and that such bias testing is audited and reported. 

Let go of assumptions: Ask yourself continuously: whom do you ask to join your team, advocate for promotion, hire, and spend time developing? How do you think your cultural background and personal experiences shape your decisions? Is anyone being left out or subjected to micro-inequities that we haven’t spotted? Are we making a recommendation based on self-interest or based on our common love of the proposal? Did we explore dissenting opinions? 

Learn and relearn relentlessly with your teams: Business leader Rene Carayol once said, “the illiterate of the 21st Century won’t be those who can’t read and write, but it will be those who can’t learn, unlearn and relearn.” One excellent resource for better understanding bias is Robin Hauser’s 2020 documentary film “Bias” which explores the pervasiveness of unconscious bias and how we can work to “de-bias our brains.”

With conscious effort we can all address the assumptions and quick judgements that may currently infuse our decision-making and evolve our leadership to a more balanced approach.