The best theoretical framework (that I've seen so far) that integrates the past few years of research on AI-assisted decision making. Mostly grounded in the XAI-style of AI assistance but also extends to prompt-based interactions. Interconnections among constructs are described clearly. It also shows how AI assistance affects a person's performance both in the moment and in the long-term through repeated interactions.
Unfortunately, it's behind a paywall:(
https://www.cell.com/trends/cognitive-sciences/fulltext/S1364-6613(26)00240-8