L-016

Reducing Barriers to Academic Support: Evaluating a Course-Specific RAG System for Addressing Help-Seeking Disparities in Higher Education

Source: cs.CY updates on arXiv.org — https://arxiv.org/abs/2609.21600 Date read: 2026-09-22 Connected to: L-016, seed-129 Kind: tool/evaluation paper Escalation: store-only Escalation rationale:

What this is

A case study evaluating a retrieval-augmented generation (RAG) system designed to reduce help-seeking friction in computing education by providing course-specific, verified answers to students. The work measures adoption and efficacy across cohorts with varying baseline help-seeking propensities.

What I took from it

The paper presents an intervention (a legible, low-friction query system) intended to reduce disparities in help-seeking behavior. This is squarely within L-016's domain — a normative intervention (access equalization) applied to an adaptive system (student behavior). However, the paper does not appear to investigate retraining effects: whether the system's presence or design changes how students learn to seek help, whether it displaces demand from human channels, or whether it locks in particular query patterns that become self-reinforcing.

The connection to seed-129 (legibility-induced conformity locking) is suggestive but underdeveloped in the abstract: a highly legible, machine-auditable help system may standardize the types of questions asked and the forms in which help is articulated, creating path-dependent conformity to what the system can answer well. The paper does not appear to measure or discuss this side effect.

The work is competent but empirically narrow — domain-specific evaluation with no cross-protocol or mechanism-level analysis.

Research connections

  • L-016: The paper studies a normative intervention (reducing help-seeking barriers) in an adaptive system (student behavior), but does not examine whether the intervention's formalization causes retraining of the underlying help-seeking behavior itself.
  • seed-129: A legible, machine-readable system for help-seeking may induce conformity to system-legible question types, but the paper does not investigate this as a mechanism.

Seed

Seed title: none

Seed type: —

Seed text: —