Reflections on AI Coaching #7 – How AI coaches could help us understand human coaching better

Although coaching research has grown significantly in the last 10 years there is still a lot we don’t know about coaching. What actually happens in a coaching session that causes change in clients? What types of coaching should we in which scenarios? How many coaching sessions are needed to achieve outcomes? Is it better to be a directive or a non-directive coach? Etc…

As a relatively new practice and academic discipline coach borrows liberally from other fields such as psychology and adult learning. Nothing wrong with that, in fact I think is it is precisely because coaching has grown organically, that we have achieved such success. (see excellent article by Prof Tatiana Bachkirova – “Old wine in new bottles: Exploring pragmatism as a philosophical framework for the discipline of coaching”). However when practice far outruns science we start walking in thin ice and scenes of snake oil salespeople start flickering on radar screens. We need to keep doing high quality research to understand coaching better, however this is not easy.

I won’t go into the details of what constitutes “good” research, but suffice to say that you would need a large number of coaches to do various coaching acts over a period of time so that one could compare outcomes. Coaches are expensive and not many typically volunteer to participate in research for free (what a pity!). There is of course the real issue of client confidentiality that prevents may coaches from participating in research. And this is where I see AI Coaches come to our aid. I have done a number of research studies where I used AI chatbot coaches to perform certain coaching tasks and I see huge potential to use this approach to boost coaching research.

It is much easier to create and use AI coaches in research than human coaches due to lower cost and availability. I am planning a number of experiments where I want to compare different coaching approaches, techniques, durations etc using AI chatbot coaches. This approach eliminates the cost and reluctance of human coaches and could dramatically speed up the time it takes to collect data. Of course we cannot completely generalised the findings of an AI coach to a human coach, but it would give us some direction of where to focus our human coaching research efforts. As I have said before, the arrival of Large Language Models (LLMs) now makes it possible for us to create very eloquent and realistic AI coaches. Very exciting!

Enough said! Let me start writing a proposal for my next AI coaching research project!