AI enfeeblement, idle supervisors and the Peter principle in agents: 2. AI enfeeblement as measured

Section 2 of 9 of the report "AI enfeeblement, idle supervisors and the Peter principle in agents" by Hiina (v1, updated 2026-10-27)
Report overview and contents: https://ai.hiina.space/reports/ai-enfeeblement-and-the-peter-principle.md
Previous: https://ai.hiina.space/reports/ai-enfeeblement-and-the-peter-principle/01-summary.md
Next: https://ai.hiina.space/reports/ai-enfeeblement-and-the-peter-principle/03-the-dark-room-problem-and-offloading.md

2. AI enfeeblement as measured

The exact experiment the question describes (train with a calculator, then test mental arithmetic without it, across training) has not been run. That gap held across about eight search wordings. The closest direct evidence says well-designed tool training preserves unaided ability, while supervised fine-tuning on tool-only traces can destroy it. Behavioral over-reliance, where models call tools they don't need, is well documented. On the human side, deskilling is real in high-stakes automation and in "practice with AI, then test without" trials, while several famous pop results are weak.

Models: tool on vs tool off

Study	Setup	Unaided ability after tool training
Toolformer, Schick 2023	GPT-J 6B, self-supervised tool fine-tune incl. calculator; evaluated with API calls disabled	No harm, slight gain (ASDiv 7.5 to 14.8); near floor, small model
TRICE, Cheng 2026	Qwen3 4B/30B thinking models, code interpreter	Final recipe keeps or improves text-only math. Ablation fine-tuned only on tool traces: text-only BeyondAIME 54.3 to 34.1, with the model writing code, inventing its output and looping. Mixing text-only traces back recovers most of it
CITI, Hao 2024	Full fine-tune of Llama-3-8B on API-Bank	Large collateral loss (GSM8K 79.5 to 26.4, HumanEval 59.2 to 0); LoRA avoids it. Ordinary catastrophic forgetting, not skill substitution
Xu & Wang 2026	RL on Countdown arithmetic with a calculator	No tool-off eval of the tool-trained model. The cheapest place to run the missing experiment

TRICE's failure mode is the most interesting data point. The model didn't just get worse; it kept the tool-shaped reasoning loop and hallucinated the tool's output when the tool was gone. That is enfeeblement as habit capture, not only skill decay.

Most tool-integrated reasoning papers (ReTool, SimpleTIR, Lin & Xu 2025) compare against a separately trained text-only model, so the "same model, tool switched off" number is almost never reported. That convention is why the gap exists. Search-RL papers likewise never report closed-book recall for the trained model.

Models: over-reliance as behavior

SMART, Qian 2025: models call tools unnecessarily over 30% of the time; Llama-3.1-8B 66% on GSM8K.
Tool-Overuse Illusion, Zeng 2026: about 0.93 unnecessary calls per simple GSM8K problem, tool use lowers accuracy on easy problems, and RL with verifiable rewards raises tool turns during training (Qwen2.5-7B 2.2 to 6.8).
Tool-Induced Myopia, Bayat 2025: with a code tool, final-answer accuracy rises but reasoning quality falls, as brute-force checks replace derivation.
Paturi 2026: merely making a related but unnecessary tool available drops the rate of answering closed questions from 98.2% to 63.5% (abstract only).
OTC-PO, Wang 2025: penalizing tool calls cuts Search-R1 calls from 3.28 to 1.04 per question with accuracy unchanged, so the extra calls were pure habit.

Models: adjacent mechanisms

RL's Razor, Shenfeld 2025: at equal new-task performance, RL forgets less than SFT, and forgetting tracks forward KL from the base model. This is the likely explanation for TRICE's RL-vs-SFT contrast and a testable prediction.
Yue et al. 2025: RLVR raises pass@1 but shrinks pass@256 below the base model. A narrowing rather than enfeeblement.
Lin et al. 2024: the RLHF alignment tax on translation and reading comprehension; weight averaging recovers much of it.
Mallen et al. 2023: retrieval rescues about 17% of PopQA answers but breaks about 10% the model already knew. The original case for using a tool only when the model's own knowledge is weak.

Humans

Finding	Design	Confidence
Bastani 2025, PNAS: plain GPT-4 practice raised practice grades 48% but exam scores without AI fell 17%; a tutor version avoided the harm	Field RCT, ~1,000 students	High
Shen & Tamkin 2026, Anthropic: engineers learning a library scored 50% with AI vs 67% without on a later quiz; delegators did worst	RCT, n=52	Medium
Liu et al. 2026: brief AI help lowered later unaided performance and persistence	RCTs, n=1,222 (abstract only)	Medium
Maier et al. 2026: AI access did not significantly change the later unaided fractions test (OR 0.80, 95% CI 0.42-1.51, so harm is not ruled out); feedback on reliance helped	Preregistered, n=704	Medium, partly null
Budzyń 2025, Lancet Gastro: adenoma detection without AI fell 28.4% to 22.4% after AI was introduced	Observational, 1,443 procedures	Medium
Casner 2014: 747 pilots kept hand-flying skills but lost cognitive navigation skills; mind-wandering 20% of thought with automation on vs 5% off, correlated with errors	Simulator, n=16	Medium
Parasuraman & Manzey 2010: automation complacency and bias affect experts and survive training	Review	High
Hembree & Dessart 1986: calculators did not hurt paper skills in K-12 except sustained use in grade 4	Meta-analysis, 79 studies (secondary summary)	Medium
Dahmani & Bohbot 2020: more GPS use predicted steeper spatial memory decline	Longitudinal n=13	Low
Woollett & Maguire 2011: taxi trainees who qualified grew posterior hippocampus	Longitudinal, self-selected	Medium (use-it side)
Sparrow 2011 "Google effect": failed replication (BF 5 for null) and in Camerer 2018	Replications	Treat as unreliable
Kosmyna 2025, "Your Brain on ChatGPT": lower EEG connectivity with LLM use	n=54, 18 in crossover; critiqued	Low
Lee et al. CHI 2025, Gerlich 2025: AI confidence correlates with less critical thinking	Self-report surveys	Low for causation

The human gap: no multi-week study randomizes LLM access and then tests unaided skill after a washout. Chooi, Lee & Li 2026 make the matching point for AI evals: disempowerment metrics measure exposure, not skill erosion.

Casner's pilot result is the best existing answer to the "what does the human do while the AI works" question: the freed attention goes to mind-wandering, and that predicts later failure. That points toward quiescence, not productive substitution.

Text licensed CC BY 4.0 (attribution: Hiina, ai.hiina.space). Source: https://github.com/hiinaspace/ai-hiina-space