Section 148 of 440
FAMILY 25: HUMAN–AI COMPLEMENTARITY
Superdomain: Technology & Augmentation. This family contains eight stable Power records.
PWR-193 · AI-assisted reasoning
Evidence route: Externally gated evidence dossier. Use the linked canonical tutorial for current teaching, accessibility, safety and editorial-review status; this evidence dossier does not enlarge that tutorial’s authority.
Summary. AI can improve selected reasoning tasks in a declared configuration, but it can also transmit error and add verification work.
Definition. Capability to develop or express ai-assisted reasoning in a declared context without inheriting broader claims.
What the current evidence supports. AI can improve selected reasoning tasks in a declared configuration, but it can also transmit error and add verification work.
Measurement boundary. Accuracy, calibration, time, verification effort and harm-weighted error versus human-only, AI-only and the better standalone baseline. This measures the declared configured system and cannot by itself establish broad transfer, unaided ability, safety or legitimacy.
Myth. AI makes a person universally smarter
Metric. Accuracy, calibration, time, verification effort and harm-weighted error versus human-only, AI-only and the better standalone baseline
Boundary. Report person, device/model/interface, task, assistance, failure state and comparator; the metric is not proof of unaided general capability.
Negative and limiting findings.
Explanations did not reliably create complementarity and incorrect advice sometimes reduced performance below the human-only baseline.
No cited evidence supports perfect, universal or consequence-free performance.
Claims this evidence cannot support.
AI makes a person universally smarter
The tool or robot's output proves an unaided, universal or permanent human superpower
A successful laboratory task proves independent real-world or clinical capability
Evidence references. [2] limiting or contrary evidence; [3] primary empirical support; limiting or contrary evidence; [9] primary empirical support; limiting or contrary evidence; [10] primary empirical support; limiting or contrary evidence; [11] primary empirical support; limiting or contrary evidence; [12] limiting or contrary evidence; [13] limiting or contrary evidence; [15] limiting or contrary evidence; official boundary context; [16] limiting or contrary evidence; official boundary context.
Next gate. External confirmation is required before higher-authority use because: a clinical, high-risk or regulated configuration.
PWR-194 · AI-assisted memory
Evidence route: Externally gated evidence dossier. Use the linked canonical tutorial for current teaching, accessibility, safety and editorial-review status; this evidence dossier does not enlarge that tutorial’s authority.
Summary. AI-supported memory is a configured retrieval capability; what is remembered by the person and what remains in the system must be reported separately.
Definition. Capability to develop or express ai-assisted memory in a declared context without inheriting broader claims.
What the current evidence supports. AI-supported memory is a configured retrieval capability; what is remembered by the person and what remains in the system must be reported separately.
Measurement boundary. Device-on retrieval accuracy, source traceability and latency plus delayed unaided recall and performance after system loss. This measures the declared configured system and cannot by itself establish broad transfer, unaided ability, safety or legitimacy.
Myth. An AI second brain permanently expands memory
Metric. Device-on retrieval accuracy, source traceability and latency plus delayed unaided recall and performance after system loss
Boundary. Report person, device/model/interface, task, assistance, failure state and comparator; the metric is not proof of unaided general capability.
Negative and limiting findings.
Expectation of external access changed what people recalled, and biased AI exposure produced later unassisted errors rather than improvement.
No cited evidence supports perfect, universal or consequence-free performance.
Claims this evidence cannot support.
An AI second brain permanently expands memory
The tool or robot's output proves an unaided, universal or permanent human superpower
A successful laboratory task proves independent real-world or clinical capability
Evidence references. [2] primary empirical support; limiting or contrary evidence; [5] limiting or contrary evidence; [6] limiting or contrary evidence; [8] primary empirical support; limiting or contrary evidence; [9] limiting or contrary evidence; [16] limiting or contrary evidence; official boundary context; [17] limiting or contrary evidence; official boundary context.
Next gate. External confirmation is required before higher-authority use because: a clinical, high-risk or regulated configuration.
PWR-195 · AI-assisted creativity
Evidence route: Externally gated evidence dossier. Use the linked canonical tutorial for current teaching, accessibility, safety and editorial-review status; this evidence dossier does not enlarge that tutorial’s authority.
Summary. AI can assist selected creative outputs, sometimes most for lower-baseline performers, while narrowing collective diversity.
Definition. Capability to develop or express ai-assisted creativity in a declared context without inheriting broader claims.
What the current evidence supports. AI can assist selected creative outputs, sometimes most for lower-baseline performers, while narrowing collective diversity.
Measurement boundary. Blinded novelty/usefulness, corpus diversity, provenance, human contribution, time and delayed unaided performance. This measures the declared configured system and cannot by itself establish broad transfer, unaided ability, safety or legitimacy.
Myth. AI unlocks unlimited personal creativity
Metric. Blinded novelty/usefulness, corpus diversity, provenance, human contribution, time and delayed unaided performance
Boundary. Report person, device/model/interface, task, assistance, failure state and comparator; the metric is not proof of unaided general capability.
Negative and limiting findings.
Assisted stories became more similar to one another, and passive reliance reduced ownership, meaning and self-efficacy.
No cited evidence supports perfect, universal or consequence-free performance.
Claims this evidence cannot support.
AI unlocks unlimited personal creativity
The tool or robot's output proves an unaided, universal or permanent human superpower
A successful laboratory task proves independent real-world or clinical capability
Evidence references. [6] primary empirical support; limiting or contrary evidence; [7] primary empirical support; limiting or contrary evidence; [16] limiting or contrary evidence; official boundary context; [17] limiting or contrary evidence; official boundary context.
Next gate. External confirmation is required before higher-authority use because: a clinical, high-risk or regulated configuration.
PWR-196 · AI-assisted perception
Evidence route: Externally gated evidence dossier. Use the linked canonical tutorial for current teaching, accessibility, safety and editorial-review status; this evidence dossier does not enlarge that tutorial’s authority.
Summary. AI can assist selected perceptual judgements, but performance belongs to the versioned reader-tool-workflow configuration.
Definition. Capability to develop or express ai-assisted perception in a declared context without inheriting broader claims.
What the current evidence supports. AI can assist selected perceptual judgements, but performance belongs to the versioned reader-tool-workflow configuration.
Measurement boundary. Sensitivity, specificity, calibration, time and severity-weighted misses versus expert-only, AI-only and combined conditions. This measures the declared configured system and cannot by itself establish broad transfer, unaided ability, safety or legitimacy.
Myth. AI gives superhuman sight
Metric. Sensitivity, specificity, calibration, time and severity-weighted misses versus expert-only, AI-only and combined conditions
Boundary. Report person, device/model/interface, task, assistance, failure state and comparator; the metric is not proof of unaided general capability.
Negative and limiting findings.
Some combined teams did not exceed the best standalone component, and incorrect advice or explanations could induce overreliance.
No cited evidence supports perfect, universal or consequence-free performance.
Claims this evidence cannot support.
AI gives superhuman sight
The tool or robot's output proves an unaided, universal or permanent human superpower
A successful laboratory task proves independent real-world or clinical capability
Evidence references. [4] primary empirical support; limiting or contrary evidence; [14] primary empirical support; limiting or contrary evidence; [17] limiting or contrary evidence; official boundary context.
Next gate. External confirmation is required before higher-authority use because: a clinical, high-risk or regulated configuration; a restricted safety or legitimacy boundary.
PWR-197 · AI-assisted decision calibration
Evidence route: Externally gated evidence dossier. Use the linked canonical tutorial for current teaching, accessibility, safety and editorial-review status; this evidence dossier does not enlarge that tutorial’s authority.
Summary. Reliance on AI can be made more selective in some interfaces, but no explanation or warning guarantees good judgement.
Definition. Capability to develop or express ai-assisted decision calibration in a declared context without inheriting broader claims.
What the current evidence supports. Reliance on AI can be made more selective in some interfaces, but no explanation or warning guarantees good judgement.
Measurement boundary. Calibration, appropriate acceptance/rejection, verification rate, workload and harm-weighted errors across advice-quality strata. This measures the declared configured system and cannot by itself establish broad transfer, unaided ability, safety or legitimacy.
Myth. A confidence score makes AI safe to trust
Metric. Calibration, appropriate acceptance/rejection, verification rate, workload and harm-weighted errors across advice-quality strata
Boundary. Report person, device/model/interface, task, assistance, failure state and comparator; the metric is not proof of unaided general capability.
Negative and limiting findings.
Incorrect AI reduced accuracy, explanations often failed to help, and acquired AI bias persisted into later unaided decisions.
No cited evidence supports perfect, universal or consequence-free performance.
Claims this evidence cannot support.
A confidence score makes AI safe to trust
The tool or robot's output proves an unaided, universal or permanent human superpower
A successful laboratory task proves independent real-world or clinical capability
Evidence references. [2] primary empirical support; limiting or contrary evidence; [3] primary empirical support; limiting or contrary evidence; [10] primary empirical support; limiting or contrary evidence; [11] primary empirical support; limiting or contrary evidence; [12] primary empirical support; limiting or contrary evidence; [13] primary empirical support; limiting or contrary evidence; [15] limiting or contrary evidence; official boundary context; [17] limiting or contrary evidence; official boundary context.
Next gate. External confirmation is required before higher-authority use because: a clinical, high-risk or regulated configuration.
PWR-198 · Agent delegation and orchestration
Evidence route: Externally gated evidence dossier. Use the linked canonical tutorial for current teaching, accessibility, safety and editorial-review status; this evidence dossier does not enlarge that tutorial’s authority.
Summary. Humans can delegate bounded tasks to agents, but the capability is the governed person-agent workflow, not autonomous output alone.
Definition. Capability to develop or express agent delegation and orchestration in a declared context without inheriting broader claims.
What the current evidence supports. Humans can delegate bounded tasks to agents, but the capability is the governed person-agent workflow, not autonomous output alone.
Measurement boundary. Task completion, error recovery, intervention rate, verification burden, authority violations and human-only/agent-only/team performance. This measures the declared configured system and cannot by itself establish broad transfer, unaided ability, safety or legitimacy.
Myth. One prompt lets an agent safely run everything
Metric. Task completion, error recovery, intervention rate, verification burden, authority violations and human-only/agent-only/team performance
Boundary. Report person, device/model/interface, task, assistance, failure state and comparator; the metric is not proof of unaided general capability.
Negative and limiting findings.
Delegation increased payoff inequality in one game, hybrid nudging did not improve group success, and participants shifted responsibility to agents.
No cited evidence supports perfect, universal or consequence-free performance.
Claims this evidence cannot support.
One prompt lets an agent safely run everything
The tool or robot's output proves an unaided, universal or permanent human superpower
A successful laboratory task proves independent real-world or clinical capability
Evidence references. [1] primary empirical support; limiting or contrary evidence; [6] primary empirical support; limiting or contrary evidence; [15] limiting or contrary evidence; official boundary context; [16] limiting or contrary evidence; official boundary context; [17] limiting or contrary evidence; official boundary context.
Next gate. External confirmation is required before higher-authority use because: a clinical, high-risk or regulated configuration.
PWR-199 · AI tutoring and adaptive learning
Evidence route: Externally gated evidence dossier. Use the linked canonical tutorial for current teaching, accessibility, safety and editorial-review status; this evidence dossier does not enlarge that tutorial’s authority.
Summary. A structured AI tutor can improve immediate learning on selected lessons; that does not prove durable or universal education benefit.
Definition. Capability to develop or express ai tutoring and adaptive learning in a declared context without inheriting broader claims.
What the current evidence supports. A structured AI tutor can improve immediate learning on selected lessons; that does not prove durable or universal education benefit.
Measurement boundary. Predeclared learning outcome, time, delayed unaided retention and unfamiliar transfer versus active human instruction and ordinary chatbot use. This measures the declared configured system and cannot by itself establish broad transfer, unaided ability, safety or legitimacy.
Myth. An AI tutor teaches every subject better than people
Metric. Predeclared learning outcome, time, delayed unaided retention and unfamiliar transfer versus active human instruction and ordinary chatbot use
Boundary. Report person, device/model/interface, task, assistance, failure state and comparator; the metric is not proof of unaided general capability.
Negative and limiting findings.
Unguided use can bypass thinking, and the positive trial relied on expert prompts, prewritten solutions and only immediate post-tests.
No cited evidence supports perfect, universal or consequence-free performance.
Claims this evidence cannot support.
An AI tutor teaches every subject better than people
The tool or robot's output proves an unaided, universal or permanent human superpower
A successful laboratory task proves independent real-world or clinical capability
Evidence references. [2] limiting or contrary evidence; [5] primary empirical support; limiting or contrary evidence; [8] limiting or contrary evidence; [16] limiting or contrary evidence; official boundary context; [17] limiting or contrary evidence; official boundary context.
Next gate. External confirmation is required before higher-authority use because: a clinical, high-risk or regulated configuration.
PWR-200 · Human–AI fallback collaboration
Evidence route: Externally gated evidence dossier. Use the linked canonical tutorial for current teaching, accessibility, safety and editorial-review status; this evidence dossier does not enlarge that tutorial’s authority.
Summary. Useful human-AI systems require tested human-only or alternative routes for model error, outage and withdrawal.
Definition. Capability to develop or express human–ai fallback collaboration in a declared context without inheriting broader claims.
What the current evidence supports. Useful human-AI systems require tested human-only or alternative routes for model error, outage and withdrawal.
Measurement boundary. Failure detection, time-to-safe-state, recovery accuracy, override success and post-outage human performance versus pre-automation baseline. This measures the declared configured system and cannot by itself establish broad transfer, unaided ability, safety or legitimacy.
Myth. A human in the loop guarantees safety
Metric. Failure detection, time-to-safe-state, recovery accuracy, override success and post-outage human performance versus pre-automation baseline
Boundary. Report person, device/model/interface, task, assistance, failure state and comparator; the metric is not proof of unaided general capability.
Negative and limiting findings.
People accepted incorrect advice, inherited bias and sometimes performed below their own baseline; explanations alone did not solve fallback.
No cited evidence supports perfect, universal or consequence-free performance.
Claims this evidence cannot support.
A human in the loop guarantees safety
The tool or robot's output proves an unaided, universal or permanent human superpower
A successful laboratory task proves independent real-world or clinical capability
Evidence references. [2] limiting or contrary evidence; [3] primary empirical support; limiting or contrary evidence; [6] primary empirical support; limiting or contrary evidence; [9] primary empirical support; limiting or contrary evidence; [10] primary empirical support; limiting or contrary evidence; [11] primary empirical support; limiting or contrary evidence; [12] primary empirical support; limiting or contrary evidence; [13] primary empirical support; limiting or contrary evidence; [15] limiting or contrary evidence; official boundary context.
Next gate. External confirmation is required before higher-authority use because: a clinical, high-risk or regulated configuration.
Family source register
Numbers in each dossier refer to this family register. A source can support one bounded proposition while simultaneously limiting transfer, certainty, generalisation or safety.
[1] Elias Fernández Domingos; Inês Terrucha; Rémi Suchon; Jelena Grujić; Juan C. Burguillo; Francisco C. Santos; Tom Lenaerts (2022). Delegation to artificial agents fosters prosocial behaviors in the collective risk dilemma. Primary research. https://doi.org/10.1038/s41598-022-11518-9
[2] Lucía Vicente; Helena Matute (2023). Humans inherit artificial intelligence biases. Primary research. https://pubmed.ncbi.nlm.nih.gov/37789032/
[3] Julia Cecil; Eva Lermer; Matthias F. C. Hudecek; Jan Sauer; Susanne Gaube (2024). Explainability does not mitigate the negative impact of incorrect AI advice in a personnel selection task. Primary research. https://pubmed.ncbi.nlm.nih.gov/38679619/
[4] Julian Senoner; Simon Schallmoser; Bernhard Kratzwald; Stefan Feuerriegel; Torbjørn Netland (2024). Explainable AI improves task performance in human-AI collaboration. Primary research. https://pubmed.ncbi.nlm.nih.gov/39730794/
[5] Greg Kestin; Kelly Miller; Anna Klales; Timothy Milbourne; Gregorio Ponti (2025). AI tutoring outperforms in-class active learning: an RCT introducing a novel research-based design in an authentic educational setting. Primary research. https://doi.org/10.1038/s41598-025-97652-6
[6] Elena Hayoung Lee; Yidan Yin; Nan Jia; Cheryl J. Wakslak (2026). Relying on AI at work reduces self-efficacy, ownership, and meaning while active collaboration mitigates the effects. Primary research. https://doi.org/10.1038/s41598-026-42312-6
[7] Anil R. Doshi; Oliver P. Hauser (2024). Generative AI enhances individual creativity but reduces the collective diversity of novel content. Primary research. https://doi.org/10.1126/sciadv.adn5290
[8] Betsy Sparrow; Jenny Liu; Daniel M. Wegner (2011). Google Effects on Memory: Cognitive Consequences of Having Information at Our Fingertips. Primary research. https://doi.org/10.1126/science.1207745
[9] Shakked Noy; Whitney Zhang (2023). Experimental evidence on the productivity effects of generative artificial intelligence. Primary research. https://doi.org/10.1126/science.adh2586
[10] Gagan Bansal; Tongshuang Wu; Joyce Zhou; Raymond Fok; Besmira Nushi; Ece Kamar; Marco Tulio Ribeiro; Daniel S. Weld (2021). Does the Whole Exceed its Parts? The Effect of AI Explanations on Complementary Team Performance. Primary research. https://doi.org/10.1145/3411764.3445717
[11] Zana Buçinca; Maja Barbara Malaya; Krzysztof Z. Gajos (2021). To Trust or to Think: Cognitive Forcing Functions Can Reduce Overreliance on AI in AI-assisted Decision-making. Primary research. https://doi.org/10.1145/3449287
[12] Ujué Agudo; Karlos G. Liberal; Miren Arrese; Helena Matute (2024). The impact of AI errors in a human-in-the-loop process. Primary research. https://doi.org/10.1186/s41235-023-00529-3
[13] Corey Lester; Brigid Rowell; Yifan Zheng; Zoe Co; Vincent Marshall; Jin Yong Kim; Qiyuan Chen; Raed Kontar; X. Jessie Yang (2025). Effect of Uncertainty-Aware AI Models on Pharmacists' Reaction Time and Decision-Making in a Web-Based Mock Medication Verification Task: Randomized Controlled Trial. Primary research. https://pubmed.ncbi.nlm.nih.gov/40249341/
[14] Jeong Hoon Lee; Ki Hwan Kim; Eun Hye Lee; Jong Seok Ahn; Jung Kyu Ryu; Young Mi Park; Gi Won Shin; Young Joong Kim; Hye Young Choi (2022). Improving the Performance of Radiologists Using Artificial Intelligence-Based Detection Support Software for Mammography: A Multi-Reader Study. Primary research. https://pubmed.ncbi.nlm.nih.gov/35434976/
[15] Elham Tabassi; National Institute of Standards and Technology (2023). Artificial Intelligence Risk Management Framework (AI RMF 1.0). Official authority. https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-ai-rmf-10
[16] C. Autio; R. Schwartz; J. Dunietz; S. Jain; M. Stanley; E. Tabassi; P. Hall; National Institute of Standards and Technology (2024). Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile. Official authority. https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence
[17] National Institute of Standards and Technology (2020). NIST Privacy Framework: A Tool for Improving Privacy Through Enterprise Risk Management, Version 1.0. Official authority. https://www.nist.gov/publications/nist-privacy-framework-tool-improving-privacy-through-enterprise-risk-management