How will YOU avoid these AI-related cognitive biases?

Email Notifications

SHARE
Request to reuse this

I'm midway through reading Jeremy Kahn's book "Mastering A.I. - A Survival Guide To Our Superpowered Future". While I find the title aspirational (can you truly master anything which is evolving as rapidly as A.I.?), the author has done a good job of providing a balanced assessment of some near and longer term benefits and risks of A.I.

What has resonated with me as it relates to project management are the following three cognitive biases:

• Automation bias - the inclination to assume that recommendations or information presented by a computer system are more accurate than that produced by a human being, even when we are presented with contradictory evidence.
• Automation neglect - the tendency to discount and ignore what a computer system is telling us, especially when it runs counter to our beliefs or desires.
• Automation surprise - the tendency to rely on computer systems and to be confused or surprised when they fail.

I've witnessed the impact of the first two biases multiple times over my career with traditional project management applications.

I've seen senior executives trust the information provided in a Project Portfolio Management solution's sexy dashboard telling them that a particular project was healthy even when the data used to populate that dashboard had undergone significant green-shifting and it was clear to any stakeholder remotely close to the project that it was on fire.

I've seen a sponsor refuse to accept a project manager's recommendation to push back a milestone date based on a Monte Carlo simulation which showed that meeting the desired date had an extremely low probability of success.

I haven't run into automation surprise yet mostly because many project management applications have the unfortunate tendency of failing regularly as the complexity or volume of data or queries increases.

In the near term, we are unlikely to fall prey to such biases when it comes to A.I.-based project management solutions. It is being well drilled into us to employ techniques such as human in the middle to verify that A.I. generated outputs are valid.

But lets fast forward a few years to when the growing pains of the current generation of A.I. tools are but distant memories.

As the reliability of the tools improves, our vigilance diminishes. The likelihood of automation bias affecting project managers, team members, and senior stakeholders will increase, especially as our ability to understand how the A.I. tools are coming to a conclusion gets harder. This will go hand-in-hand with automation surprise. When A.I. tools fail, we might lack the experience or knowledge to understand how to troubleshoot it and if we have become too reliant on the tool doing what we would have done manually in the past, our ability to take over might have atrophied.

The impacts of automation neglect are likely to remain fairly constant. For stakeholders who have a preconceived belief that they don't wish to have challenged, a high confidence contrary answer from a more reliable A.I. is unlikely to sway them. Mandating that users are required to follow the A.I.'s guidance is not the solution as it just increases the potential impacts of automation bias and automation surprise.

So as you contemplate your future as a project manager, what will YOU do to reduce the impacts of these biases as A.I.-enabled project management continues to mature?