In the context of [[human-AI creative collaboration]], I believe an Al co-creative system can help the user get into a state of [[flow]]. [[Mihaly Csikszentmihalyi]] proposes that [[Conditions of flow]]: - That the task sits at the right spot of difficulty: not too challenging, not too hard - That it provides clear feedback - That it has a well defined goal How can we design human-AI co-creative systems to provide these conditions? One idea is that AI can help fill the skill gap in difficult activities and move it closer to the sweet spot. Providing feedback on the activity could also be done by a model, via AI to human communication. This is also relevant in a dialogic model of interaction because [[AI to human communication is critical for the AI to be considered a partner and not a tool]] A good example is Codex, [[Evaluating Large Language Models Trained on Code (Codex paper)]]. The user requires less syntax knowledge for tackling a challenge of the same size. Additionally, the person's mental energy can be focused on the bigger creative picture rather than in the syntactic details. I have personally used Codex for that purpose and I feel I can achieve more in less time and with less effort. I feel as if my own creativity is augmented. From: