Page - 23 - in Coaching im digitalen Wandel

The challenges of coaching and mentoring 23 basic level. Coachbots – algorithms that attempt to emulate the questioning approach of coaches are still very clunky, easily flummoxed and unlikely to be taken for a human. The investment to take them to that point would be enor- mous, requiring at the least a library of tens of thousands of coaching or men- toring conversations. Even with such a resource to draw upon, the challenge remains in determining what »good« looks like. If the AI »learns« from analys- ing thousands of poor or mediocre conversations, that is what it will emulate. What an AI lacks includes the ability to draw on wider life experience, the associated skill of intuition and the ability to connect at the level of shared val- ues, which underlies the creation of a professional friendship. A big question about the last of these is the extent to which clients want to build a professional friendship with a therapist or coach. Sometimes it can be easier (as the USC researchers have found) to disclose in a neutral environment – in a human to human relationship, there is always at least a residual discomfort arising from the fear of being judged. It is also unclear whether the AI performs well at all the functions of a trauma therapist or just at the diagnostic functions. Who is under threat? So, what is to stop Ellie’s near-future cousins replacing coaches? The answer, I suspect, lies in the purpose of coaching and in the coaches’ relative maturity. In terms of coaching purpose, a well-accepted framework defines four types of intervention, of increasing complexity and capability required by the coach: skills, performance, behaviour change and transformational. Both skills and performance coaching function primarily through the application (conscious or unconscious) of a relatively small number of algorithms. (The much-overused »on a scale of 0 to 10« process is one example of such an algorithm.) As long as there are frequently repeated patterns, seen in multiple situations and clients, an AI can learn to do these at least as well as a human. The USC researchers have already, for example, developed an AI for the US military that can coach soldiers to negotiate with an Afghan war criminal – a skill that requires, among other things, a high level of cultural sensitivity (Gratch, 2014). So, skills and performance coaching are most at risk. Behavioural and trans- formational coaching are less at risk, because the core of the process is the internal reflection that clients undertake, much of it outside the formal coach- ing sessions. The personal experience and wisdom of the coach (or mentor) is more relevant, as is the art of helping someone connect with their values, find the metaphors that release their dreams, develop intrinsic motivation or let go. Open-Access-Publikation im Sinne der CC-Lizenz BY-NC-ND 4.0