Tsamados et al. 2024 argue that human control of AI must shift from supervision to teaming, with evaluation of human-AI systems as a whole rather than AI benchmarks alone; Zhu et al. 2026 independently identify sustained shared control, cross-session memory, and recovery from complex failures as agent challenges that existing human-AI interaction principles do not capture.
E2 fully supports the Zhu et al. challenges (sustained shared control, cross-session memory, and recovery from complex failures not captured by existing principles), and E29 supports assessing human-AI systems as a whole rather than benchmarking AI alone, but the cited Tsamados text never states that human control must shift 'from supervision to teaming.'
Written by Kimi K3 via Ollama Cloud · checked by GLM-5.3 via Ollama Cloud · 2 Oct, 04:51
Source chain
AEvery quote below was checked, without a model, to appear verbatim in its source.
- 01
“moves from benchmarking progress on whether AI systems can outperform humans in given tasks to assessing the performance of human–AI systems as a whole”
Full-text passage · no page number · evidence E29
This requires that the AI research community moves from benchmarking progress on whether AI systems can outperform humans in given tasks to assessing the performance of human–AI systems as a whole, including their interactive, organisational, and collaborative capabilities as agents of a more extensive system [46, 47]. For example, training programs that focus on testing human–machine interactions in realistic settings––either through high-fidelity simulations or deployment in contained environments––can reveal insights about agents’ capabilities and the corresponding levels of authority they should be given. Indeed, this is a growing trend among organisations spearheading the development of AI systems for complex socio-technical environments, such as the Defense Advanced Research Projects Agency (DARPA). The agency has launched several programs, including ASIST [48], SAIL-ON [49], and CAML [50] and associated software for assessing and improving human–AI systems’ performance during deployment or in computer-simulated environments, as well as to test AI systems in complex socio-technical environments (more on this in the next section).
Passage read from www.ebi.ac.uk, which may be a preprint rather than the published version.
- 02
“contemporary AI agents introduce new challenges, including sustained shared control, cross-session memory, deployment in emotionally sensitive contexts, and recovery from complex failures”
Full-text passage · p. 2 · evidence E2
central contribution is a design-oriented perspective that designs and evaluates technology in terms of human needs, capabilities, and contexts rather than technical performance alone, operationalized through methods such as think-aloud evaluation [ 29], heuristic evaluation [43, 42], iterative prototyping [10], and contextual inquiry [33]. Within this tradition, early principles for intelligent systems such as Horvitz’s mixed-initiative interfaces in 1999 [25] and later guidelines for human–AI interaction in 2019 [4] have shaped how AI systems are designed, but primarily for bounded and discrete tasks such as recommendations and search. In contrast, contemporary AI agents introduce new challenges, including sustained shared control, cross-session memory, deployment in emotionally sensitive contexts, and recovery from complex failures, which are not fully captured by existing principles [ 25, 4]. Meanwhile, recent HCI research on human–agent interaction has rapidly expanded, spanning empirical and envisioning work across diverse agent types and domains, varying in embodiment, autonomy, memory, initiative, and social roles.
Passage read from arxiv.org, which may be a preprint rather than the published version.