Back

Expectation management in humans and LLMs

Menashe, B.; Drake, A.; Ben-Shachar, M.

2025-03-11 neuroscience
10.1101/2025.03.09.642185 bioRxiv
Show abstract

Mirative markers, such as "surprisingly", explicitly encode a violation of expectations. Such markers are used for expectation management during communication. Sensitivity to mirative markers relies on two abilities: i) updating expectations upon recognizing a mirative marker, and ii) identifying expectation violations warranting the use of a mirative marker. In this study, we compared sensitivity to mirative markers in humans and large language models (LLMs). In part 1, we used a sentence-completion task, where humans and LLMs were presented with sentence fragments and asked to continue them. Results show that for both humans and LLMs, the presence of a mirative marker significantly increased response entropy and decreased top-response probability, in line with theoretical accounts of mirative processing. In part 2, we created a novel task, mirative polarity selection, where humans and LLMs are presented with a sentence pair and asked to select whether it was connected by a mirative marker ("surprisingly") or an anti-mirative marker ("unsurprisingly"). Results show that LLMs perform at an impressive human-equivalent level. We conclude that both humans and LLMs use mirative markers as cues for calibrating their subsequent expectations during sentence processing.

Matching journals

The top 4 journals account for 50% of the predicted probability mass.

1
Cognition
44 papers in training set
Top 0.1%
22.3%
2
PLOS ONE
4510 papers in training set
Top 19%
10.0%
3
Philosophical Transactions of the Royal Society B
51 papers in training set
Top 0.3%
9.0%
4
Scientific Reports
3102 papers in training set
Top 8%
9.0%
50% of probability mass above
5
PLOS Computational Biology
1633 papers in training set
Top 6%
6.2%
6
Journal of Cognitive Neuroscience
119 papers in training set
Top 0.3%
4.3%
7
Trends in Hearing
12 papers in training set
Top 0.1%
2.6%
8
eneuro
389 papers in training set
Top 4%
2.1%
9
Neurobiology of Language
28 papers in training set
Top 0.1%
1.8%
10
Journal of Neurophysiology
263 papers in training set
Top 0.4%
1.7%
11
iScience
1063 papers in training set
Top 18%
1.5%
12
Proceedings of the National Academy of Sciences
2130 papers in training set
Top 35%
1.5%
13
Journal of Vision
92 papers in training set
Top 0.3%
1.5%
14
The Journal of Neuroscience
928 papers in training set
Top 6%
1.3%
15
Behavior Research Methods
25 papers in training set
Top 0.1%
1.3%
16
Journal of Speech, Language, and Hearing Research
10 papers in training set
Top 0.1%
1.2%
17
Frontiers in Neuroscience
223 papers in training set
Top 5%
1.2%
18
Frontiers in Psychology
49 papers in training set
Top 0.8%
1.2%
19
Attention, Perception, & Psychophysics
17 papers in training set
Top 0.1%
1.1%
20
Frontiers in Human Neuroscience
67 papers in training set
Top 2%
0.9%
21
Journal of The Royal Society Interface
189 papers in training set
Top 4%
0.9%
22
Journal of Neural Engineering
197 papers in training set
Top 2%
0.9%
23
Psychonomic Bulletin & Review
14 papers in training set
Top 0.1%
0.9%
24
The Journal of the Acoustical Society of America
33 papers in training set
Top 0.2%
0.9%
25
Frontiers in Psychiatry
83 papers in training set
Top 3%
0.8%
26
Hearing Research
49 papers in training set
Top 0.3%
0.8%
27
Psychological Review
19 papers in training set
Top 0.2%
0.7%
28
Consciousness and Cognition
17 papers in training set
Top 0.4%
0.6%
29
BMC Bioinformatics
383 papers in training set
Top 8%
0.6%