Unpacking human–AI collaboration: conceptualisations and emerging research streams

Despite the growing adoption of AI across organisations, fragmented knowledge of key concepts and measurement approaches constrains systematic evaluation and cumulative progress. This review aims to harmonise the conceptual vocabulary of HAIC, map how conceptual framings align with evaluation emphases, and surface an stream structure from co-occurrence patterns, thereby supporting more comparable HAIC assessment designs. We reviewed 149 peer-reviewed studies (2018–2025) from AISeL, ACM, and IEEE using a concept-centric approach. We identified 16 recurring themes for coding and mapping, 10 of which we further defined by synthesising consistent definitional passages to specify boundary conditions that distinguish collaboration from tool use and automation. To identify concept–evaluation link patterns, we computed a co-occurrence matrix and a Jaccard-normalised overlap indicator. Our findings reveal four preliminary streams that organise these linkages. Overall, the study proposes a harmonised cross-disciplinary vocabulary and an empirically grounded mapping that strengthens interpretability and comparability in HAIC assessment.