On June 13th, we held the TENTH session of the "Coaching meets AI" dialogue series. We further delved into the development of a prompt script (i.e., a sequence of relevant prompts) for triadic AI coaching that utilizes the capabilities of GPT (without text modules).
Specifically, we focused on the second coaching phase, i.e., the analysis of the existing coaching issue.
The key insights were that the prompting strategy here should involve: • Initially asking the AI to identify the underlying causes of the coaching issue based on the provided information. • On this basis, follow-up questions can then be asked. These may include: o Requesting the AI to explain one of the causes it mentioned in more detail, o and/or asking the AI to identify the causes that might be behind one of the causes it identified. • Alternatively, or following these follow-up questions, it may be useful to first ask the coachees and then the AI to create a ranking of the different causes • and/or to determine the causal relationships between the various causes, thereby creating an overall causal-analytical picture of the causes underlying the current coaching issue.
The central task for the coaches in this context is to keep the coachees in focus regarding these different prompting options and to give them enough space and creative room. In other words, care must be taken not to delegate too much guidance to the AI.
Another important insight was that using ChatGPT without text modules and with coaching diagnostic knowledge is generally not so simple, at least not when it comes to problem analysis. Because when you ask ChatGPT to analyze the underlying causes of the existing coaching issue based on the previously provided information, it usually delivers relatively general answers. Nevertheless, these can still provide valuable suggestions for the coaching process.
Against this background, Annette Jäckel will focus on the capability of pattern recognition and bring her corresponding experiences to the next session on June 27th.