ABCDEFGHIJKLMNOPQRSTUVWXYZ
1
#TitleAuthorsYearSourceLink/DOIKey FindingsSample & SubjectsAI SettingsEcological Validity
2
1Your Brain on ChatGPT: Accumulation of Cognitive DebtKosmyna, Hauptmann, Yuan et al.2025arXiv (MIT Media Lab)https://arxiv.org/abs/2506.08872EEG showed ChatGPT users had lowest brain connectivity (~47% reduction). 83% couldn't quote from own AI-assisted essays. Progressive disengagement over 4 months.N=54 (ages 18-39), Boston-area university studentsDefault. Standard ChatGPT web interface, no custom instructions specified.Yes — reflects typical student ChatGPT use. Lab-controlled but naturalistic task. Not peer-reviewed, small sample.
3
2AI Tools in Society: Impacts on Cognitive Offloading and Critical ThinkingGerlich, M.2025Societies, 15(1), 6https://www.mdpi.com/2075-4698/15/1/6Strong negative correlation (r=-0.75) between AI usage and critical thinking, mediated by cognitive offloading (r=+0.72). Younger participants (17-25) showed highest dependency.N=666 surveys + 50 interviews; diverse UK adultsDefault (self-reported). Survey of natural AI tool use habits. No experimental manipulation.Yes — directly measures natural AI use in daily life. High ecological validity.
4
3ChatGPT as a Cognitive Crutch: Evidence from RCT on Knowledge RetentionBarcaui, A.2025Social Sciences & Humanities Open, 12https://www.sciencedirect.com/science/article/pii/S2590291125010186AI-assisted students scored 57.5% vs 68.5% on surprise retention test 45 days later (Cohen's d=0.68). AI group studied 3.2 hrs vs 5.8 hrs.N=120 undergraduate business students (18-24), BrazilDefault. GPT-4 free web interface as study aid. No custom prompts specified.Yes — students used standard ChatGPT exactly as any student would.
5
4Cognitive Ease at a Cost: LLMs Reduce Mental Effort but Compromise DepthStadler, Bannert & Sailer2024Computers in Human Behavior, 160https://www.sciencedirect.com/science/article/pii/S0747563224002541ChatGPT users reported significantly lower cognitive load but produced lower-quality reasoning and weaker argumentation than search engine users.N=91 university students, randomly assigned to ChatGPT-3.5 or GoogleDefault. Standard ChatGPT-3.5 interface. No custom system prompts described.Yes — students used ChatGPT-3.5 as any user would for research.
6
5Beware of Metacognitive Laziness: Effects of GenAI on LearningFan, Tang, Le et al.2025British Journal of Educational Technology, 56https://arxiv.org/abs/2412.09315ChatGPT group outperformed in essay scores but showed no advantage in knowledge gain or transfer. Coined "metacognitive laziness" — AI users engaged less in self-monitoring.Randomized experiment; 4 groups (ChatGPT, human expert, writing analytics, control)Default. Standard ChatGPT interface for learning support. No custom prompts.Yes — reflects typical unstructured ChatGPT use for studying.
7
6Google Effects on Memory: A Meta-Analytical ReviewGong & Yang2024Frontiers (PMC)https://pmc.ncbi.nlm.nih.gov/articles/PMC10830778/Meta-analysis confirmed the "Google effect" — people forget information they can look up — tied to cognitive load and self-esteem. Effect stronger on mobile.Meta-analysis (multiple studies)N/A - meta-analysis of search engine studies, pre-LLM era.Partial - foundational for digital cognitive offloading; extends to AI by analogy.
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100