Agentic education · Learning science

Actionable companion · Feedback & scaffolding

Learning with agency: From AI dependence to reflective collaboration: psychological ownership, competency anxiety, and perceived support in AI-assisted learning

AI-assisted editorial implementation guide, separate from the original research. Proposed actions are not claims or recommendations attributed to the source authors. No independent peer review of this guide is asserted.

Read original paperSearch paper agentCC BY 4.0

Evidence and context

Title and abstract screening; stage labels describe reported settings, not verified numerical ages. Design labels require full-text appraisal. Reported stage(s): Higher education. Verify the participant population and task before transferring the findings.

Original authors’ abstract, reproduced under the linked source license:

Abstract Generative artificial intelligence (GenAI) can improve efficiency in academic work. However, it also raises a central psychological question: when AI helps produce a task, why do some learners still experience the final work as their own, whereas others experience it as polished but psychologically distant? This study examines psychological ownership and competency anxiety in AI-assisted learning, distinguishing between dependent AI outsourcing and reflective human–AI collaboration. An exploratory qualitative interview-based design with supplementary background information was used. The background form was used only for sample description and interview preparation, followed by in-depth interviews with 50 undergraduate and postgraduate students from five Chinese universities, interviews with 10 faculty members and administrators, and 128 student critical incident records. Data were analyzed through a hybrid deductive–inductive thematic analysis, with NVivo 14 used as a supporting tool for code management and matrix comparison. In participants’ accounts, dependent AI outsourcing was associated with weaker psychological ownership, described in terms of reduced cognitive and authorial presence during planning, reasoning, and revision. This weakened sense of ownership was, in turn, linked to shallow memory, difficulty explaining an individual’s work, reduced meaning-making, and stronger competency anxiety. Reflective human–AI collaboration, by contrast, was associated with retained ownership when students planned before using GenAI, evaluated AI output, revised it in their own voice, and could explain their final decisions. Perceived educational support, including feedback, clear AI-use guidance, and psychologically safe learning environments, was described as helping students treat AI use as a learnable practice rather than a hidden shortcut. Faculty and administrators also framed these patterns as issues of assessment feasibility, workload, policy clarity, and curriculum design, rather than solely as student choice. Because the data were collected at five Chinese universities, the proposed model is presented as context-sensitive: the observed dynamics may be amplified where assessment is strongly product-oriented, formative feedback is constrained, and AI use is difficult to discuss openly. This study contributes to psychological ownership theory by extending it to human–AI interaction and proposing that, in this setting, ownership depends on cognitive and authorial presence rather than solely on task completion. Keywords: AI-assisted learning, cognitive offloading, competency anxiety, generative artificial intelligence, human-AI collaboration, perceived educational support, psychological ownership, self-regulated learning

What kind of evidence is this?

Supporting AI learning science · Qualitative / mixed methods

Stage: Higher education. Screening labels describe reported context and study-design language. They are not a risk-of-bias rating or proof of effectiveness for an age band.

Study methods · source passage

This study employed an exploratory qualitative interview-based design with supplementary background information (Levitt et al., 2018; Tong et al., 2007). A brief background information form was administered first to record minimal categorical contextual information: age range, sex, academic level, broad discipline, frequency of GenAI use, main AI-supported task types, and willingness to discuss a recent AI-assisted learning episode. These items were used only for sample description and interview preparation; they were not treated as scale scores, covariates, or variables for cross-case statistical comparison. The qualitative analysis comprised two connected strands: a learner-experience strand based on student interviews and critical incident records, and a stakeholder-context strand based on faculty and administrator interviews. The strands were first examined separately and then compared for convergence, divergence, and contextual interpretation (Palinkas et al., 2015).

Limitations and uncertainty · source passage

This study has limitations. First, it was conducted across five universities within a single Chinese higher-education context. The discussion identifies ways in which product-oriented assessment, feedback constraints, instructor authority, policy ambiguity, and the perceived safety of disclosure may have shaped the observed dynamics. However, this study cannot determine which features are culturally specific and which reflect more general assessment configurations. The findings should not be generalized to Chinese higher education as a whole, and cross-national comparisons should match task type, class size, feedback design, and AI-policy conditions rather than using country as a proxy for culture. Future research should include multinational comparative studies, including samples from Latin America, Europe, and South Asia, to examine whether similar ownership and competency-anxiety dynamics appear under different educational cultures and institutional policy environments. Second, discipline patterns were interpreted through task outsourcing affordance, but the sample was not homogeneous across STEM, humanities/social sciences, and arts-related fields. Future work should balance disciplinary distribution where possible and should systematically sample matched task types within each discipline. This would allow researchers to test whether generatability, process visibility, and explanation requirements predict ownership more strongly than broad disciplinary labels. Third, this study is an exploratory qualitative interview-based study supported by supplementary background information. The contrasts reported in the findings are interpretive summaries rather than scale scores or inferential statistical tests, and the propositions require quantitative, longitudinal, or experimental testing.

Design the system before the pilot

Use this study to inform a learning mechanism or implementation question. It is not direct evidence that autonomous agents improve learning. Keep the proposed agent behavior separate from what the source actually tested.

  • Give feedback on the learner’s reasoning, then fade hints. Record whether the learner can complete the next task without that scaffold.
  • Use plan–attempt–check–reflect prompts and compare confidence with actual performance. Keep learner choice and reflection visible.
  • Use common rubrics, blind scoring where feasible and a fresh unaided reasoning task. Document permissible assistance and avoid treating AI-detector scores as proof of misconduct.

A 90-day district pilot

  1. Days 1–15 · Appraise. Have the curriculum lead and a teacher read the methods and limitations. Record the study design, sample, comparator, outcomes and evidence gaps. Define a local question that the study can help inform.
  2. Days 16–30 · Design. Choose one disciplinary task. Require a first attempt, a documented AI consultation, a critique of the output and an independent explanation. Align the comparison group and scoring rubric with the same objective. Name an accountable owner and set the baseline, success threshold and budget before implementation.
  3. Days 31–60 · Test. Start with a small voluntary cohort, preserve an accessible non-AI route and require teacher review. Use synthetic or de-identified examples; approve data handling and tool terms through district processes.
  4. Days 61–90 · Decide. Blind-score reasoning and accuracy on a new task without AI. Distinguish model examination scores, student perceptions and actual learner gains; only the last directly tests educational benefit. Compare with baseline, document uncertainty and stop or revise if learning worsens, access is unequal or critical errors persist.

The board decision brief

Report the learning problem, evidence quality, participating population, total cost including review time, measured benefit, errors, access differences and remaining uncertainties. Make a dated recommendation to stop, revise or expand. A positive perception survey alone does not establish learning improvement.

Attribution and reuse

Xiaoling Li; Xujin Xian; Peitao Du. From AI dependence to reflective collaboration: psychological ownership, competency anxiety, and perceived support in AI-assisted learning. Frontiers in Psychology (2026-07-15). https://doi.org/10.3389/fpsyg.2026.1903784

Companion prepared 2026-10-30. Original authorship and DOI are preserved. The abstract is an attributed source extract; the implementation plan is new AI-assisted editorial material, offered under the same CC BY 4.0 license. This guide has not undergone independent peer review.