Some Ways To Handle Cognitive Biases
The below table summarises the main cognitive biases; with some ways to handle them
Principal bias (definition) |
Related bias |
What is it |
How to handle |
Confirmation (failing to consider alternatives, ignoring dissenting views, going for an easy option, etc) |
- Affected heuristic - Group think - Self-interest bias - Authority bias |
- Strong preference for one answer - Failing to look at alternatives and anti-thesis - Not approaching dissenting views with an open mind |
- Constructive confrontation - Dialectic argument, thesis, anti-thesis, synthesis - Team composition (to carry diversity, eg views, background. etc - Role-playing |
Anchoring (blind preference for initial data/pattern) |
- Saliency bias - Neglect of reversal to mean/ compounding |
- Assuming an initial data range or increased/decreased pattern is a whole range that will continue (mental numerical stickiness) | - Develop alternative hypothesis (anti-thesis) - Hard-core data gathering & analysis (including scenario planning) - What are the team processes |
Loss aversion (failure to ignore costs already spent, ie sunk, or any asymmetrical valuing of losses and gains) |
- Sunk cost fallacy - Endowment effect - Account book loss fear |
- Mental numerical stickiness or any kind of asymmetric value of losses or gains | - Clear forward-looking mindset - good analytical tools (NPV analysis valuing options, Bayesian thinking) |
Availability (preference for existing mental maps &/or influenced by most recent facts or events) |
- Substitution bias - Overemphasis given to current events |
- Use the map or story we have at hand rather than investing in understanding the new complex situation | - Value of additional information - Explicit options to wait vs to decide now |
Over-optimism (shown by overconfidence, illusion of control, failure to contemplate negative outcomes) |
- Overconfidence - Illusion of control - Disaster neglect |
- Underestimating low probability events | - Explicit extreme downside case - Pre-mortems |
(source: Charles Conn et al 2018)