EMCC is proud to announce the first individual accreditation of an artificial intelligence at Foundation level of coaching. PF3, a coachbot developed by the University of Witwatersrand is not the first AI to pass the Turing Test (which assesses whether a human interacting with an AI can distinguish it from another human), but it is the first in the world of coaching.
PF3 outperformed 40% of coaches, in a controlled experiment. The test was made tougher by the fact that all the clients in the experiment were trained coaches themselves, so expected to be better able to decide if the coaching (which was delivered by email to remove any other clues of identity) was given by a human or not. Each client was coached twice - once by a human and once by PF3 - and asked to rate both whether the online coach was human and the quality of the coaching overall.
Overall, the clients were correct in their identification of AI versus human only 47% of the time. Rating the quality of the coaching, based on criteria such as being non-judgemental and asking powerful questions, the clients scored the AI close to the median of the best and worst human coaches.
Further tests posed ethical dilemmas to the same panel of human coaches and PF3, with the average ratings for the AI's responses being assessed as slightly higher (3% on average) than those for the humans and more consistent (10%).
PF3's software designers maintain that it continues to improve with each coaching session it conducts and the AI's analysis of its own learning journey was part of the evidence submitted to the accreditation panel.
