{"id":"0d1242d0-a77a-40a9-852f-e5f99539c3b6","arxiv_id":"2411.12761","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":3.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper argues for a shift to human-AI joint research and introduces the ART, ARA, and ARP collaboration models, without supplying new empirical evidence.","lead":"This paper proposes that human research should become human-AI joint research, structured by two paradigms and three AI roles: tool, assistant, and participant. A generalist might read it to see how AI is being recast from a passive instrument into an active collaborator in brain and social science research.","discovery_kind":"review","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The ARP section conflates 'AI as experimental stimulus/subject' with 'AI as autonomous co-researcher'; every cited example is the former, so the claim that it is 'now' time to treat AI as an independent research participant lacks evidential support.","rationale":"The reader's conditional verdict is appropriate. My stress-test sharpens the reason: the most novel component (ARP) is not merely under-supported by future-looking speculation; it equivocates on the word 'participant'. The cited studies all involve AI as an experimental agent whose behavior is manipulated to observe human reactions, not as an epistemic contributor that shapes research questions, designs, or interpretations. This distinction is checkable by reclassification. Since the paper is a position piece, this does not falsify the proposal as a research agenda, but it does mean the 'now' in 'it is high time' is not established. The appropriate disposition remains CONDITIONAL: the authors should either provide examples where AI contributes autonomously to at least one core research step, or explicitly soften ARP to 'AI as simulated participant/stimulus' and separate that from AI as co-researcher. The reader's weakest assumption is in the same neighborhood; my concern is more specific about the equivocation in the cited evidence, hence partial agreement.","tokens_in":11278,"tokens_out":4079,"duration_ms":41944,"concrete_test":"Audit every ARP citation in the 'AI as Research Participant' section using a two-question rubric: (1) Does the AI determine or modify the research question, hypothesis, experimental design, or data interpretation, rather than merely being a condition or stimulus? (2) Could the same study be run with a scripted human confederate or an inert stimulus with identical scientific content? If all citations answer 'no' to (1) and 'yes' to (2), the ARP concept reduces to AI-as-subject or AI-as-stimulus, and the central claim is unsupported. Publish the audit table and re-run the argument using only the examples that survive the rubric.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim has two parts: (a) transition to human-AI joint research is due 'now', and (b) AI should be an independent entity actively participating in research. Both rest on Section 'AI as Research Participant Empowers Human Research'. The section defines ARP as AI 'actively engag[ing] in research process, contributing autonomously', but all four supporting studies treat AI as a manipulated agent or stimulus inside a human experiment: Mahmoodi et al. (2022) uses AI partners to study human conformity; Dell'Acqua et al. (2023) uses an AI player in a cooking game; Traeger et al. (2020) uses a 'vulnerable robot' to shape team conversation; Pataranutaporn et al. (2023) primes beliefs about a chatbot. In none of these does AI set the research question, design the experiment, analyze the data, or interpret the results. The term 'research participant' is thus equivocal: a participant-as-subject or participant-as-stimulus is not a participant-as-collaborator. The normative leap 'it should now be recognized as an independent entity' is load-bearing and unsupported by the paper's own evidence. This is not a disagreement with the consensus that AI is useful; it is an inference gap between the cited demonstrations and the proposed role.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper argues that AI has moved from being a mere instrument to an active collaborator in scientific research, and that it is time for researchers to transition to human-AI joint research. It reviews literature on human cognition, emotion, and collaborative learning, then proposes two new research paradigms—the AI-Brain Science Research Paradigm and the AI-Social Sciences Research Paradigm—together with three human-AI collaboration models: AI as a research tool (ART), AI as a research assistant (ARA), and AI as a research participant (ARP). The paper also sketches research methods, including empirical studies and questionnaire surveys, and claims these paradigms will reshape future research practice.","tokens_in":11462,"tokens_out":5004,"duration_ms":48453,"significance":"If the proposed framework were operationalized and supported with direct evidence, it could serve as a useful organizing structure for studying human-AI collaboration in brain science and social sciences. The paper has value as a broad literature review and as an explicit articulation of the tool/assistant/participant trichotomy, which many researchers discuss implicitly. However, the central claim that AI is already ready to act as an independent research participant is not supported by the cited evidence, and the proposed paradigms and methods lack operational definitions. The contribution is therefore currently more of a position statement or agenda than a validated framework. The paper does not provide machine-checked proofs, reproducible code, parameter-free derivations, or falsifiable predictions; its strengths lie in synthesis and conceptual organization rather than in new empirical or formal results.","major_comments":[{"comment":"The definition of ARP states that AI 'actively engages in research process, contributing autonomously,' but all four cited supporting studies—Mahmoodi et al. (2022), Dell'Acqua et al. (2023), Traeger et al. (2020), and Pataranutaporn et al. (2023)—treat AI as an experimental stimulus or a manipulated agent within human experiments. In none of these does AI set the research question, design the experiment, analyze the data, or interpret the results. The term 'research participant' is therefore equivocal: a participant-as-subject or participant-as-stimulus is not a participant-as-collaborator. The normative conclusion that AI 'should now be recognized as an independent entity actively participating in research' is load-bearing and unsupported by the paper's own evidence. The authors should either redefine ARP to the weaker claim that AI can serve as a simulated participant or subject, adjusting the paper's conclusions accordingly, or provide direct evidence of AI performing autonomous research functions in a co-researcher role.","section":"AI as Research Participant Empowers Human Research"},{"comment":"This section claims to outline methods for conducting human-AI joint research, but it only summarizes two existing studies—Doshi and Hauser (2023) on AI-assisted story writing and Jia and Tu (2024) on AI and critical thinking—and does not present a concrete methodological protocol for implementing the proposed ART/ARA/ARP models. No operational definitions, procedures, or validation criteria are provided for how researchers would assign roles, collect data, or measure the contribution of AI as a research participant. As a result, the promised 'practical approaches for integrating AI as a full research partner' are not delivered. The section should either be expanded with explicit methodological guidance or relabeled as an illustrative review of relevant existing studies.","section":"Research Methods of Human-AI Joint Research"},{"comment":"The two 'new research paradigms' are under-specified. The text only states that they are based on traditional paradigms and 'consider the impact of AI' on cognition/emotion or social interaction, without articulating their assumptions, scope, or distinctive methodological commitments. It is not made clear what is genuinely new about these paradigms relative to existing applications of AI in brain science and social sciences. In particular, the paper cites Hardy et al. (2023), which already distinguishes AI as tools, models, and participants, but does not discuss how the proposed ART/ARA/ARP taxonomy relates to or extends that prior framework. The authors should explicitly define each paradigm, state its novel elements, and explain how the three collaboration models instantiate these paradigms in testable ways.","section":"New Research Paradigms of Human-AI Joint Research"}],"minor_comments":[{"comment":"The sentence 'As artificial general intelligence (AGI) AGI advances' contains a duplicated 'AGI'; the phrase 'these capabilities are critical for AI to effectively engage' is also vague and should specify which capabilities are meant and how they would be assessed.","section":"Introduction"},{"comment":"The author name is spelled 'Dell'Aqua' in the text but 'Dell'Acqua' in the reference list; please make the spelling consistent. In addition, the reference to Han et al. (2024) is incomplete, listing 'LC, R.' as an author without a full name.","section":"AI as Research Participant Empowers Human Research"},{"comment":"The labels 'human-AI joint thinking' and 'AI identifies emerging research trends and gap' in the figure are not defined in the text; either define these terms explicitly or remove them from the figure for clarity.","section":"Figure 2"},{"comment":"The citation 'Van den Boosche et al. (2006)' appears to be a typo for 'Van den Bossche et al.'; please verify the correct spelling against the reference list.","section":"Fundamentals of Human Research"},{"comment":"The paper lacks a dedicated 'Limitations and Future Work' section. Given that the proposed paradigms and methods are not empirically validated in this manuscript, a candid discussion of limitations would strengthen the paper's scholarly credibility.","section":"Conclusion"},{"comment":"Given that Hardy et al. (2023) explicitly distinguishes AI as tools, models, and participants, the absence of a direct comparison between that taxonomy and the proposed ART/ARA/ARP trichotomy weakens the novelty claim and should be addressed.","section":"Related Work"}],"recommendation":"major_revision","confidential_remarks":"This manuscript is best viewed as a vision or position paper rather than an empirical or methodological contribution. The main risk is overclaiming the current readiness of AI to serve as an independent research participant; the paper's own cited evidence supports the weaker claim that AI can act as a simulated participant or stimulus in human experiments. I would suggest that the editor consider whether the journal's scope accommodates such position papers, and if so, require the authors to substantially reframe the central claim as a research agenda with explicit testable predictions and operational definitions. There is no indication of misconduct; however, the novelty overlap with Hardy et al. (2023) should be carefully checked during revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth a look if you work on human-AI collaboration. The paper gives a clean tripartite taxonomy – AI as tool, assistant, participant – and maps it onto two research paradigms for brain science and social sciences. That is genuinely useful as a shared vocabulary, and the literature review is broad and accurate. The writing is clear; the figures help.\n\nWhat's new is mostly the packaging. The ART/ARA/ARP split is close to Hardy et al.'s tools/models/participants, which the paper cites but does not engage with on that point. So the novelty is thin.\n\nThe bigger problem is the ARP section. The paper defines AI as a research participant as 'actively engaging in the research process, contributing autonomously,' but every example it gives is AI as a manipulated agent or stimulus inside a human experiment – AI partners in a conformity task, an AI player in a cooking game, a vulnerable robot shaping team dynamics, a chatbot whose perceived intentions are primed. None of those involve AI setting the question, designing the study, analyzing data, or interpreting results. So the claim that 'it should now be recognized as an independent entity actively participating in research' is an inference gap, not a conclusion the cited literature supports. The paper never addresses that gap.\n\nThe methods section is really a summary of two existing studies (Doshi & Hauser; Jia & Tu), not a new method. That's fine for a position piece, but it means there is no new empirical content.\n\nWho is this for? Someone looking for a compact review of AI roles in HCI and social-science research, and a vocabulary for talking about them. If the venue wants a rigorous argument, the ARP claim needs a proper defense – either independent evidence of genuine autonomous contribution, or a clear statement that ARP is aspirational rather than present practice.\n\nI'd send it to peer review, but I'd expect the referees to push back on the ARP conflation and the uncredited overlap with Hardy. With revision it could be a modestly useful position paper; as it stands the central claim overreaches.","headline":"A clear position paper with a useful vocabulary, but the central ARP claim leans on cited studies that only treat AI as stimulus or subject, not as an autonomous co-researcher.","tokens_in":12089,"tokens_out":2048,"would_cite":false,"duration_ms":20931,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper argues that AI should now be treated as an independent research participant, and it proposes two paradigms and three collaboration models to move scientific research toward human-AI joint inquiry.","keywords":["human-AI joint research","AI research participant","AI research assistant","AI research tool","research paradigm","brain science","social sciences","human-AI collaboration"],"falsifier":"A controlled comparison in which independent experts evaluate the novelty and correctness of research contributions produced by human-only teams versus teams with AI participants, and then track whether the AI-inclusive findings replicate, would settle whether AI as research participant adds scientific value.","tokens_in":10975,"feed_emoji":"🤖","tokens_out":3652,"duration_ms":36501,"temperature":0.7,"pith_summary":"The paper argues that artificial intelligence has outgrown the role of a passive instrument and should be treated as an independent research collaborator. Its central claim is that researchers should move from human-only research to human-AI joint research, and it supplies two paradigms for that transition, one grounded in brain science and one in social science. If the claim holds, AI systems with cognitive and social capabilities would take part in experiments, surveys, and theory-building alongside human researchers, and research methods would be redesigned around that partnership. The paper's contribution is a structured vocabulary and scaffold for making that shift, not yet a proof that AI works as a participant.","feed_headline":"AI should co-run research, not just assist humans","feed_subtitle":"Paper proposes brain-science and social-science paradigms where AI acts as tool, assistant, or independent participant.","key_machinery":"The load-bearing structure is a three-level division of AI's role in research: tool, assistant, and participant. The participant tier carries the argument, because that is where AI stops being subordinate to human direction and enters the research as an autonomous collaborator. The paper grounds this tier in published findings that AI influences human neural decision-making, group behavior, trust, and perceptions of reliability, then uses the tool-assistant-participant ladder to map AI onto each stage of the research process.","core_discovery":"The paper diagnoses the current state of scientific research as still human-centered even as AI plays an expanding role, and it claims the time has come to formalize AI's place in the research process. It proposes two new research paradigms, the AI-Brain Science Research Paradigm and the AI-Social Sciences Research Paradigm, and within them three models of human-AI collaboration: AI as a research tool (ART), AI as a research assistant (ARA), and AI as a research participant (ARP). The strongest claim is normative: because AI 'continues to develop human-like capabilities,' it should now be recognized as an independent entity actively participating in research, not merely a subordinate apparatus.","pith_inferences":["If taken literally, AI-as-participant would require rethinking scholarly authorship and credit, though the paper does not spell out those rules.","The line between assistant and participant is not sharply drawn; a testable extension would define observable criteria, such as autonomy in choosing hypotheses, to classify AI roles.","The existing empirical examples are mostly short-term lab effects, so extending ARP to long-horizon discovery would require measuring cumulative research output rather than in-session influence.","The paper itself notes that AI's full research potential is not yet realized and that a comprehensive framework has been lacking, so the proposal is best read as a roadmap rather than a demonstrated outcome."],"forward_implications":["Research designs would need to register AI as a participant, affecting how studies are planned, reported, and evaluated for validity.","Brain-science studies would examine human-AI interaction with the same instruments used for human-human interaction, such as fMRI and EEG.","Social-science surveys and experiments would treat AI avatars and conversational agents as social actors whose presence measurably changes behavior and trust.","The ART-ARA-ARP distinction gives researchers a shared language for specifying what role AI plays in a given study.","Human-AI joint research methods would combine empirical experiments on joint creativity with questionnaire surveys on critical thinking.","If taken literally, AI-as-participant would require rethinking scholarly authorship and credit, though the paper does not spell out those rules.","The line between assistant and participant is not sharply drawn; a testable extension would define observable criteria, such as autonomy in choosing hypotheses, to classify AI roles.","The paper itself notes that AI's full research potential is not yet realized and that a comprehensive framework has been lacking, so the proposal is best read as a roadmap rather than a demonstrated outcome."],"supporting_citations":[{"why":"Supplies fMRI evidence that humans conform to AI partners during decision-making, activating the dorsal anterior cingulate cortex, which supports the AI-as-participant claim.","marker":"Mahmoodi et al. 2022"},{"why":"Shows in a multiplayer game that AI outperforming humans reduces human performance and trust, evidence that AI participants alter group dynamics.","marker":"Dell'Acqua, Kogut, and Perkowski 2023"},{"why":"Demonstrates that beliefs about a chatbot's cognitive and emotional capacities shape perceived trustworthiness, supporting the ARP model.","marker":"Pataranutaporn et al. 2023"},{"why":"Shows that vulnerable robots positively shape human conversational dynamics, supporting AI's role as a social participant.","marker":"Traeger et al. 2020"},{"why":"Argues that AI can transform social science research paradigms, motivating the proposed AI-Social Sciences Research Paradigm.","marker":"Grossmann et al. 2023"},{"why":"Provides the integrated view of cognition and emotion that underpins the paper's account of human research fundamentals.","marker":"Pessoa 2013"},{"why":"Empirical study showing generative AI enhances story-writing creativity, an example used for the human-AI joint creative thinking method.","marker":"Doshi and Hauser 2023"},{"why":"Questionnaire study of 637 college students linking AI use to increased self-efficacy and critical thinking, illustrating the survey-based method.","marker":"Jia and Tu 2024"}],"fun_headline_variants":["AI as research participant: new paradigms for brain and social science","Human-AI joint research: AI as tool, assistant, or participant","AI should join research as a participant, not just a tool","Two new paradigms put AI inside the research process"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The framework assumes that AI will acquire the cognitive and social capabilities needed to act as a genuine research participant, and that research involving AI participants will yield scientifically valid results.","fun_headline_variants_meta":{"raw":{"variants":["AI as research participant: new paradigms for brain and social science","Human-AI joint research: AI as tool, assistant, or participant","AI should join research as a participant, not just a tool","Two new paradigms put AI inside the research process"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000152,"raw_usage":{"total_tokens":1141,"prompt_tokens":820,"completion_tokens":321,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":436,"completion_tokens_details":{"reasoning_tokens":251}},"tokens_in":436,"tokens_out":321,"duration_ms":4548,"temperature":1.0,"reasoning_tokens":251,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T19:19:19.658788+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A controlled comparison in which independent experts evaluate the novelty and correctness of research contributions produced by human-only teams versus teams with AI participants, and then track whether the AI-inclusive findings replicate, would settle whether AI as research participant adds scientific value.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies fMRI evidence that humans conform to AI partners during decision-making, activating the dorsal anterior cingulate cortex, which supports the AI-as-participant claim."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Shows in a multiplayer game that AI outperforming humans reduces human performance and trust, evidence that AI participants alter group dynamics."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Demonstrates that beliefs about a chatbot's cognitive and emotional capacities shape perceived trustworthiness, supporting the ARP model."},{"cited_title":"L.; Strohkorb Sebo, S.; Jung, M.; Scassellati, B.; and Christakis, N","cited_arxiv_id":null,"evidence_quote":"Shows that vulnerable robots positively shape human conversational dynamics, supporting AI's role as a social participant."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the integrated view of cognition and emotion that underpins the paper's account of human research fundamentals."},{"cited_title":"R.; and Hauser, O","cited_arxiv_id":null,"evidence_quote":"Empirical study showing generative AI enhances story-writing creativity, an example used for the human-AI joint creative thinking method."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Questionnaire study of 637 college students linking AI use to increased self-efficacy and critical thinking, illustrating the survey-based method."}],"review_version":1}