Cognitive Debt in the GenAI Era - Listening to Practitioners' Voices on the Loss of Understanding

Theocharis Tavantzis; Bianca Trinkenreich; Michaela Greiler; Júlia Condé Araújo; Marina Condé Araújo; Italo Santos; Alexander Serebrenik; Margaret-Anne Storey; Maria Teresa Baldassarre

arXiv · 2610.37888

Cognitive Debt in the GenAI Era - Listening to Practitioners' Voices on the Loss of Understanding

Theocharis Tavantzis·Bianca Trinkenreich·Michaela Greiler·Júlia Condé Araújo·Marina Condé Araújo·Italo Santos·Alexander Serebrenik·Margaret-Anne Storey·Maria Teresa Baldassarre

View original source ↗Read source document ↗Export metadata ↓

Abstract

Fast-paced AI-assisted software development risks reducing developers' understanding of the systems they build. This growing gap between what the system does and what developers understand about it, described as ``cognitive debt'', is not new, but AI has exacerbated it. This has led to a recent surge in blog posts and online discussions, in which practitioners recognize and write about cognitive debt, and share their experiences well before academic studies can capture them. To understand their concerns and inform future research and practice, we conduct a grey literature review of 76 sources, focusing on how cognitive debt is conceptualized and manifests, its contributing factors, its consequences, and the practices to prevent, detect, and mitigate it. We find that cognitive debt can accumulate unnoticed behind working code and passing tests, particularly when developers accept AI-generated solutions without building understanding. Practitioners report consequences such as harder maintenance and debugging, fragmented knowledge in the team, and reduced productivity. They propose practices such as documenting design rationale, review-sized changes, comprehension gates, and knowledge-sharing rituals, along with signals like reliability metrics to detect it. Our findings give software teams a structured view of emerging practical approaches in AI-assisted development and stress the need to preserve human understanding alongside faster code production.

Explore related subjects

cs.SE

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

Theocharis Tavantzis, Bianca Trinkenreich, Michaela Greiler, Júlia Condé Araújo, Marina Condé Araújo, Italo Santos, Alexander Serebrenik, Margaret-Anne Storey, Maria Teresa Baldassarre. 2026-10-27. Cognitive Debt in the GenAI Era - Listening to Practitioners' Voices on the Loss of Understanding. https://arxiv.org/abs/2610.37888

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

Cloud-OpsBench: A Reproducible Benchmark for Agentic Root Cause Analysis in Cloud Systems

LLM agents are explored for automating root cause analysis (RCA) in cloud-native systems, creating a need to evaluate both diagnostic correctness and the quality of the supporting investigation. Existing static benchmarks offer repeatable inputs but limited system-facing interaction, while live testbeds expose tools but hinder controlled comparison because incident evidence varies across runs; both paradigms focus primarily on final answers. To address these limitations, we present Cloud-OpsBench, an evaluation infrastructure for interactive and evidence-grounded cloud RCA. It comprises 754 runtime-verified cases across 57 fault types on two microservice workloads spanning application services and Kubernetes platform layers. Each fault is captured as a state snapshot and replayed through standard diagnostic interfaces, with outcome labels and diagnostic evidence graphs that enable matched comparisons and process-level analysis. Evaluating eleven LLMs on Cloud-OpsBench reveals that strong outcome accuracy often masks incomplete investigations: while Joint RCA Accuracy (JRA) reaches up to 0.78 on OnlineBoutique and 0.68 on TrainTicket, models identify the correct root cause without achieving evidence closure (e.g., in 33.1% and 29.4% of evaluated cases for GPT-5). These results demonstrate that final-answer accuracy alone cannot establish whether an agent has traced the failure and collected the evidence needed to confirm its root cause.

cs.SE↗

arXiv2026

MCP Server Architecture Patterns for LLM-Integrated Applications

The Model Context Protocol (MCP), introduced by Anthropic in November 2024, defines a standardized interface for connecting large language models (LLMs) to external tools, data sources, and services. Within months of release, hundreds of community-built MCP servers appeared on GitHub, but no software-maintenance literature has yet described how the ecosystem is being structured in production. This industry experience paper catalogues five recurring MCP server architectural patterns observed across an enumerated corpus of fifteen independently developed servers (five production servers from the ANSYR voice AI platform plus ten public servers from the official MCP registry): Resource Gateway, Tool Orchestrator, Stateful Session Server, Proxy Aggregator, and Domain-Specific Adapter. Each pattern is described in the structured form established by Gamma et al.: context, problem, solution, and consequences. We also document four anti-patterns and a set of cross-cutting concerns around authentication, versioning, and observability. Quantitative evaluation contributes two measurements: inter-rater reliability of the taxonomy across two independent LLM raters on 54 held-out servers (Cohen's kappa = 0.76), which also localizes three pattern-boundary ambiguities; transport overhead measured end-to-end on loopback (stdio: 0.01 ms p50; streamable-http: 0.39 ms p50) and modeled for cross-host paths from same-region network baselines (~30 ms p50 baseline plus protocol overhead). Version 2 withdraws the tool-count analysis of version 1, whose numbers could not be traced to production records. Code, corpus, and prompts are released at https://github.com/rodriguescarson/mcp-patterns-icsme2026.

cs.SE↗

arXiv2026

TATG: Tracking-Aware Testing Objective for LLM-based Test Generation

Complex Java methods challenge automated unit test generation, as achieving high coverage and fault detection requires satisfying intricate testing requirements involving program states, execution paths, and assertions. Existing LLM-based approaches leverage program context, static analysis, and dynamic feedback but lack explicit tracking of individual testing requirements across iterations, potentially leading to redundant generation and unresolved testing objectives. We present TATG, a tracking-aware LLM-based unit test generation framework that unifies structural coverage and mutation-based fault detection. TATG introduces a structured objective representation capturing testing requirements derived from static analysis and dynamic feedback, enabling fine-grained tracking of their satisfaction status. It employs a two-stage workflow that first improves structural coverage and subsequently strengthens assertions through mutation-guided hardening. We evaluate TATG on 141 complex Java methods against state-of-the-art baselines, including KTester and PANTA. Results show average improvements of 22.15, 20.14, and 37.66 percentage points in line coverage, branch coverage, and mutation score, respectively. TATG also outperforms general-purpose coding assistants, including Codex and Claude Code, under the same open-weight model configurations, achieving higher average coverage and mutation scores while generating fewer tests and consuming fewer tokens. These findings highlight the value of explicitly tracking testing objectives for effective LLM-based test generation.

cs.SE↗