The Absent Cognitive Baseline: Theorizing a Structural Gap in AI-Native College Students' Academic Self-Assessment

Synthesis: Jia & Xu (2026) propose the Absent Cognitive Baseline (ACB) as a conceptual framework — a theory-building contribution to the knowledge base's foundational strand. Students entering college in 2026–27 are among the first cohorts to have completed most of their secondary schooling under pervasive generative AI availability. ACB names the possible structural gap that results when sustained substitutive AI use during the formative high-school window reduces the independent cognitive encounters on which academic Self-Assessment depends. The construct spans three dimensions — the unknowability of one's own cognitive boundary, false calibration from AI-generated fluency, and the de-normalization of cognitive struggle — and a conceptual model identifies three groups of moderating variables (use-, learner-, and environment-level) shaping whether ACB emerges.

Key Findings

1. ACB is a structural, not individual, condition. The gap follows from the conditions under which learning experience accumulates, not from any learner's effort or ability. "A student does not reach an absent baseline through carelessness — the baseline is absent because the conditions that would have built it were displaced." ACB operates as a gradient, not a binary: the more formative academic experience is mediated substitutively, the fewer independent reference points exist for self-calibration.

2. Three dimensions of ACB. The construct manifests across three analytically distinct planes that may co-occur or interact:

   • Unknowability of one's own cognitive boundary (epistemic plane): the learner cannot reliably determine where their own ability ends and where AI-generated capacity begins.
   • False calibration of the sense of understanding (phenomenological plane): a subjective sense of understanding grounded in the fluency of AI output rather than the learner's own cognitive representation.
   • De-normalization of cognitive struggle (normative-interpretive plane): struggle loses its status as a meaningful learning signal and is reinterpreted as inefficiency.

3. Three groups of moderating variables. ACB is most likely to emerge when three conditions co-occur: substitutive and frequent AI use during the formative years, limited prior knowledge and underdeveloped self-regulated learning, and an environment providing little structured guidance for AI use. Use-level moderators include the mode of interaction (substitutive vs. complementary) and frequency; learner-level include prior knowledge (strong learners benefit, weak learners are harmed) and SRL capacity; environment-level include guided vs. unguided AI architectures.

4. Distinct from adjacent constructs. ACB differs from Cognitive Offloading (a momentary strategy; ACB is the cumulative condition that may result from extended offloading), metacognitive laziness (a process during AI use; ACB persists even when AI is absent), and the Dunning-Kruger effect (which predicts overconfidence; a student with ACB lacks the experiential record to ground any Self-Assessment, whether confident or cautious).

5. The central shift in analytical level. Most AI in Education research asks whether AI improves or harms performance. ACB asks whether sustained AI use during formative years may alter a learner's ability to assess their own learning: "generative AI may change not only what students can do academically, but the conditions under which they come to know what they can do."

Theoretical foundations

The paper grounds ACB in three converging research strands. Metacognitive theory (Flavell; Bjork) establishes that self-assessment depends on experiential reference points built from accumulated experience. Self-regulated learning research establishes that these reference points form during adolescence and are far from automatic — microanalytic studies show many U.S. high school students graduate without consolidated metacognitive monitoring skills, before AI entered the picture. Post-2022 AI-education research (Bastani, Fan, Lehmann) establishes that generative AI can reduce the cognitive encounters on which those reference points depend. ACB is the construct that connects these strands, giving the combined pattern a name, a structure, and conditions under which it may emerge.

The paper is careful to distinguish AI availability from AI substitution — a student who used AI to check grammar while doing the intellectual work independently may have accumulated substantial cognitive experience, while one who routinely prompted AI to produce finished essays may not. It also notes that "AI-native" marks a shared historical condition, not a shared competence, deliberately avoiding the failed "digital native" thesis (Prensky) and its assumption of uniform fluency.

The three dimensions in detail

• Dimension One — Unknowability of the cognitive boundary: In traditional environments, students develop a rough sense of their limits through repeated encounters with tasks they can and cannot complete. When a large portion of academic output is AI-produced or shaped, the learner has fewer data points from which to infer the contours of their own ability.
• Dimension Two — False calibration: Processing fluency routinely misleads judgment of learning (Bjork et al. 2013). In AI-saturated environments this shifts qualitatively: fluency is delivered as a property of the AI's output, not generated by the learner's engagement. A student who reads a well-structured AI explanation may feel they understand, but the feeling reflects the received text, not the state of their own cognitive representation — a metacognitive miscalibration.
• Dimension Three — De-normalization of struggle: With pervasive AI, unresolved cognitive difficulty decreases because the tool resolves it before the student fully engages. The risk is that struggle becomes read as inefficiency rather than a normal feature of learning — a weak baseline that had limited opportunity to develop.

Plausible feedback paths connect the dimensions: sustained absence of struggle may reduce the independent encounters needed to map one's boundaries; a persistent false sense of understanding may reduce motivation to seek struggle.

Empirical indicators and methodological implications

While ACB cannot be reduced to a single score, the paper identifies indicators: students who cannot describe their own writing or reasoning ability absent AI; who conflate AI-assisted output quality with their own competence; who respond to no-AI tasks with disproportionate uncertainty, anxiety, or surprise; and who interpret cognitive struggle as inefficiency or failure. These are entry points for inquiry, not diagnostic criteria.

Methodologically, the paper argues that existing frameworks reach "ACB's door but do not enter" — performance-focused studies document what happens to outcomes under AI but not what learning feels like from the learner's perspective. It proposes interpretative phenomenological analysis (IPA) as a methodologically coherent path into the subjective layer ACB theorizes, clarifying whether the construct corresponds to something real in learners' lived experience. The moderating-variable model yields separately falsifiable predictive claims (e.g., cross-cohort comparisons of calibration error), designed to be refined or refuted by empirical encounter.

What this means for practice

• Learners. Treat a felt sense of understanding after reading fluent AI output as unreliable evidence of your own grasp, because that fluency is a property of the generated text rather than of your cognitive representation.
• Instructors. Create structured opportunities for sustained independent cognitive work — the pedagogical response the paper names if the missing reference point is the problem — rather than treating AI-assisted and independent work as interchangeable.
• Instructors. Separate availability from substitution: a student who used AI to check grammar while doing the thinking independently accumulated cognitive experience, while one who prompted AI for finished work may not have.
• Researchers. Adopt interpretative phenomenological analysis with purposively sampled learners, because performance-focused studies document what happens to outcomes but not what learning feels like from the learner's perspective.
• Administrators. Resist blanket indictment and blanket endorsement alike: the construct is proposed for a definable subset — sustained substitutive use during formative years in systems with unrestricted AI access — and the paper deliberately prescribes no fixes before the phenomenon is understood.

Limitations

• It is a conceptual, theory-building paper with no participants, measures, or data: the three dimensions and the moderating-variable model are analytical constructs, and the paper states ACB "cannot be reduced to a single score."
• The claim is prospective and untested: it concerns students entering college in 2026–27, so no cohort has yet been followed from secondary school into higher education.
• The proposed evidence is a design, not a result: purposive sampling along AI-use frequency, mode, and depth, two rounds of semi-structured interviews plus reflective journals, and IPA analysis are outlined as future work.
• The listed indicators (inability to describe one's own ability absent AI, conflating AI output quality with competence, disproportionate uncertainty on no-AI tasks, reading struggle as inefficiency) are described as entry points for inquiry rather than diagnostic criteria, and the construct's distinctness from cognitive offloading, metacognitive laziness, and the Dunning-Kruger effect is argued rather than measured.

Citation

Jia, Y., & Xu, J. (2026). The Absent Cognitive Baseline: Theorizing a Structural Gap in AI-Native College Students' Academic Self-Assessment. EdArXiv Preprint. (Westcliff University)