We will test whether AI assistance that performs developing cognitive operations weakens adolescents’ later independent judgment, and develop evidence-based safeguards that keep human reasoning meaningfully in the loop.
We will test whether AI assistance that performs developing cognitive operations weakens adolescents’ later independent judgment, and develop evidence-based safeguards that keep human reasoning meaningfully in the loop.
This is an independent research project by Ma Lin, publishing as Salah ibn Musa. It investigates how generative-AI assistance affects adolescents’ developing capacity for evidence evaluation, reasoning, and independent judgment.
The project will first develop and validate a research battery for measuring independent reasoning, transfer to new problems, delayed performance, and the depth of cognitive delegation. It will then conduct a staged pilot study comparing conventional scaffolded AI assistance with a developmentally recalibrated assistance policy.
The recalibrated policy will use periodic evidence of a learner’s independent performance to determine whether an AI system should provide a cue, scaffold the reasoning process, participate in the task, or refrain from performing the developing cognitive operation.
The initial phase will be conducted only after appropriate ethics review, site permissions, guardian consent, and adolescent assent. The project will begin with a feasible pilot sample and expand only if the measurement tools, recruitment process, and safeguarding procedures are validated.
The study will report favorable, unfavorable, null, and inconclusive results. Its practical aim is to produce evidence-based safeguards for adolescent-facing AI systems that preserve human reasoning, verification, transfer, and meaningful oversight.
The project addresses a neglected human-factors layer of AI safety. As advanced AI systems become better at performing reasoning and judgment, users may achieve higher assisted performance while losing opportunities to exercise, verify, and retain the underlying capacities. In adolescents, this may weaken the human judgment and oversight capabilities on which safe use of advanced systems depends.
The causal chain is as follows. First, the study will measure whether delegating developing cognitive operations to AI changes later independent evidence evaluation, reasoning, and transfer. Second, it will test whether reduced independent capacity or difficulty detecting flawed AI advice is associated with delegation itself rather than only with task difficulty or reduced practice. Third, it will test periodic AI-restricted assessments as a governance signal. Finally, validated findings will be translated into safeguards that guide AI systems to cue, scaffold, co-perform, or refrain according to demonstrated human capacity and uncertainty.
The expected contribution is a measurable safeguard against a pathway by which increasingly capable AI could produce dependent or over-trusting human decision-makers. The project does not claim to solve catastrophic risk directly. It aims to preserve human verification, judgment, transfer, and meaningful oversight, and to publish evidence-bounded guidance for safer adolescent-facing AI systems.
Team Member
所申请的资金将支持分阶段的研究项目，而非立即开展大规模研究。
最低资金将支持测量测试的开发与验证、伦理与场地许可准备、预注册、参与者保护、飞行员招募、参与者报酬、安全数据管理和初步分析。
理想的资金支持规模大幅提升的试点、重复独立绩效评估、延迟转移测量、盲测结果评分、研究协助、统计分析以及结果的公开发布。
该项目不会声称在必要仪器和程序得到验证前完成完整的大规模青少年研究。任何扩展都将取决于伦理批准、招聘可行性、数据质量和可用资金。
资金不会用于无关的商业活动。所有涉及未成年人的研究都需适当的审查、监护人同意、青少年同意、隐私保护，以及对阳性、阴性、无效和不确定结果的透明报告。
The requested funding will support a staged research program rather than an immediate large-scale study.
The minimum funding will support the development and validation of the measurement battery, ethics and site-permission preparation, preregistration, participant safeguarding, pilot recruitment, participant compensation, secure data management, and preliminary analysis.
The ideal funding will support a substantially larger pilot, repeated independent-performance assessments, delayed-transfer measurement, blinded outcome scoring, research assistance, statistical analysis, and open dissemination of the results.
The project will not claim to complete a full large-scale adolescent study before the necessary instruments and procedures have been validated. Any expansion will depend on ethics approval, recruitment feasibility, data quality, and available funding.
The requested funding will support a staged research program rather than an immediate large-scale study.
The minimum funding will support the development and validation of the measurement battery, ethics and site-permission preparation, preregistration, participant safeguarding, pilot recruitment, participant compensation, secure data management, and preliminary analysis.
The ideal funding will support a substantially larger pilot, repeated independent-performance assessments, delayed-transfer measurement, blinded outcome scoring, research assistance, statistical analysis, and open dissemination of the results.
The project will not claim to complete a full large-scale adolescent study before the necessary instruments and procedures have been validated. Any expansion will depend on ethics approval, recruitment feasibility, data quality, and available funding.
This is an independent research project by Ma Lin, publishing as Salah ibn Musa, studying how AI assistance affects adolescents’ developing capacity for evidence evaluation and reasoning.
The project will compare conventional scaffolded assistance with a developmentally recalibrated policy that uses periodic AI-restricted evidence of a learner’s independent performance to determine how much of the same cognitive operation AI may perform. Approximately 300 adolescents aged 14–17 would be studied only after ethics review, school or site permissions, guardian consent, and adolescent assent.
The study will measure immediate performance, independent no-AI capacity, transfer to new problems, delayed performance, and the depth of cognitive delegation. It will report positive, negative, null, and inconclusive findings.
The goal is to produce evidence and practical safeguards for adolescent-facing AI systems: when an AI should cue, scaffold, co-perform, or refrain from completing a developing cognitive operation.
The project addresses a neglected human-factors layer of AI safety. As advanced AI systems become better at performing reasoning and judgment, users may achieve higher assisted performance while losing opportunities to exercise, verify, and retain the underlying capacities. In adolescents, this may weaken the human judgment and oversight capabilities on which safe use of advanced systems depends.
The causal chain is:
Measure whether delegation of specific developing cognitive operations changes later independent evidence evaluation and reasoning.
Identify whether reduced independent capacity, failure to detect flawed AI advice, or loss of transfer is caused by delegation itself rather than by task difficulty, reduced practice, or generic AI exposure.
Test a governance signal: periodic AI-restricted evidence of independent performance.
Translate validated findings into safeguards that govern AI behavior—cue, scaffold, co-perform, or refrain—according to demonstrated human capacity and uncertainty.
Publish the methods, null findings, and evidence-bounded design guidance openly so other researchers and developers can evaluate and improve them.
The expected safety contribution is not a claim that this study will solve catastrophic risk. It is a measurable safeguard against a pathway by which increasingly capable AI could create dependent, over-trusting, or less capable human decision-makers. Preserving human verification, transfer, and judgment supports meaningful human oversight and reduces the chance that advanced AI assistance is treated as authoritative when the user can no longer independently evaluate it.
The requested funding will support a staged research program rather than an immediate large-scale study.
The minimum funding will support the development and validation of the measurement battery, ethics and site-permission preparation, preregistration, participant safeguarding, pilot recruitment, participant compensation, secure data management, and preliminary analysis.
The ideal funding will support a substantially larger pilot, repeated independent-performance assessments, delayed-transfer measurement, blinded outcome scoring, research assistance, statistical analysis, and open dissemination of the results.
The project will not claim to complete a full large-scale adolescent study before the necessary instruments and procedures have been validated. Any expansion will depend on ethics approval, recruitment feasibility, data quality, and available funding.
No funds will be used for unrelated commercial activity. All research involving minors will require appropriate review, guardian consent, adolescent assent, privacy protection, and transparent reporting of positive, negative, null, and inconclusive findings.
No comments yet. Be the first to share your thoughts.