GPT is an effective tool for multilingual psychological text analysis

Опубликована: Май 19, 2023

The social and behavioral sciences have been increasingly using automated text analysis to measure psychological constructs in text. We explore whether GPT, the large-language model underlying artificial intelligence chatbot ChatGPT, can be used as a tool for several languages. Across 15 datasets (n = 47,925 manually annotated tweets news headlines), we tested different versions of GPT (3.5 Turbo, 4, 4 Turbo) accurately detect (sentiment, discrete emotions, offensiveness, moral foundations) across 12 found that (r 0.59-0.77) performs much better than English-language dictionary 0.20-0.30) at detecting judged by manual annotators. nearly well as, sometimes than, top-performing fine-tuned machine learning models. Moreover, GPT’s performance has improved successive model, particularly lesser-spoken Overall, may superior many existing methods analysis, since it achieves relatively high accuracy languages, requires no training data, is easy use with simple prompts (e.g., “is this negative?”) little coding experience. provide sample code video tutorial analyzing application programming interface. argue other models democratize making advanced natural language processing capabilities more accessible, help facilitate cross-linguistic research understudied

Язык: Английский

---

A manifesto for applying behavioural science

Nature Human Behaviour, Год журнала: 2023, Номер 7(3), С. 310 - 322

Опубликована: Март 20, 2023

Язык: Английский

---

GPT is an effective tool for multilingual psychological text analysis

Proceedings of the National Academy of Sciences, Год журнала: 2024, Номер 121(34)

Опубликована: Авг. 12, 2024

The social and behavioral sciences have been increasingly using automated text analysis to measure psychological constructs in text. We explore whether GPT, the large-language model (LLM) underlying AI chatbot ChatGPT, can be used as a tool for several languages. Across 15 datasets ( n = 47,925 manually annotated tweets news headlines), we tested different versions of GPT (3.5 Turbo, 4, 4 Turbo) accurately detect (sentiment, discrete emotions, offensiveness, moral foundations) across 12 found that r 0.59 0.77) performed much better than English-language dictionary 0.20 0.30) at detecting judged by manual annotators. nearly well as, sometimes than, top-performing fine-tuned machine learning models. Moreover, GPT’s performance improved successive model, particularly lesser-spoken languages, became less expensive. Overall, may superior many existing methods analysis, since it achieves relatively high accuracy requires no training data, is easy use with simple prompts (e.g., “is this negative?”) little coding experience. provide sample code video tutorial analyzing application programming interface. argue other LLMs help democratize making advanced natural language processing capabilities more accessible, facilitate cross-linguistic research understudied

Язык: Английский

---

Which Humans?

Опубликована: Сен. 22, 2023

Large language models (LLMs) have recently made vast advances in both generating and analyzing textual data. Technical reports often compare LLMs’ outputs with “human” performance on various tests. Here, we ask, “Which humans?” Much of the existing literature largely ignores fact that humans are a cultural species substantial psychological diversity around globe is not fully captured by data which current LLMs been trained. We show responses to measures an outlier compared large-scale cross-cultural data, their cognitive tasks most resembles people from Western, Educated, Industrialized, Rich, Democratic (WEIRD) societies but declines rapidly as move away these populations (r = -.70). Ignoring human machine psychology raises numerous scientific ethical issues. close discussing ways mitigate WEIRD bias future generations generative models.

Язык: Английский

---

Spectro-temporal acoustical markers differentiate speech from song across cultures

Nature Communications, Год журнала: 2024, Номер 15(1)

Опубликована: Июнь 6, 2024

Abstract Humans produce two forms of cognitively complex vocalizations: speech and song. It is debated whether these differ based primarily on culturally specific, learned features, or if acoustical features can reliably distinguish them. We study the spectro-temporal modulation patterns vocalizations produced by 369 people living in 21 urban, rural, small-scale societies across six continents. Specific ranges spectral temporal modulations, overlapping within categories societies, significantly differentiate from Machine-learning classification shows that this effect cross-culturally robust, being classified solely their all societies. Listeners unfamiliar with cultures classify using similar cues as machine learning algorithm. Finally, are better able to discriminate song than a broad range other variables, suggesting modulation—a key feature auditory neuronal tuning—accounts for fundamental difference between categories.

Язык: Английский

---

Large language models predict human sensory judgments across six modalities

Scientific Reports, Год журнала: 2024, Номер 14(1)

Опубликована: Сен. 13, 2024

Язык: Английский

---

Globally, songs and instrumental melodies are slower and higher and use more stable pitches than speech: A Registered Report

Science Advances, Год журнала: 2024, Номер 10(20)

Опубликована: Май 15, 2024

Both music and language are found in all known human societies, yet no studies have compared similarities differences between song, speech, instrumental on a global scale. In this Registered Report, we analyzed two datasets: (i) 300 annotated audio recordings representing matched sets of traditional songs, recited lyrics, conversational melodies from our 75 coauthors speaking 55 languages; (ii) 418 previously published adult-directed song speech 209 individuals 16 languages. Of six preregistered predictions, five were strongly supported: Relative to songs use higher pitch, slower temporal rate, (iii) more stable pitches, while both used similar (iv) pitch interval size (v) timbral brightness. Exploratory analyses suggest that features vary along "musi-linguistic" continuum when including lyrics. Our study provides strong empirical evidence cross-cultural regularities speech.

Язык: Английский

---

Perils and opportunities in using large language models in psychological research

PNAS Nexus, Год журнала: 2024, Номер 3(7)

Опубликована: Июнь 28, 2024

The emergence of large language models (LLMs) has sparked considerable interest in their potential application psychological research, mainly as a model the human psyche or general text-analysis tool. However, trend using LLMs without sufficient attention to limitations and risks, which we rhetorically refer "GPTology", can be detrimental given easy access such ChatGPT. Beyond existing guidelines, investigate current limitations, ethical implications, specifically for show concrete impact various empirical studies. Our results highlight importance recognizing global diversity, cautioning against treating (especially zero-shot settings) universal solutions text analysis, developing transparent, open methods address LLMs' opaque nature reliable, reproducible, robust inference from AI-generated data. Acknowledging utility task automation, annotation, expand our understanding psychology, argue diversifying samples expanding psychology's methodological toolbox promote an inclusive, generalizable science, countering homogenization, over-reliance on LLMs.

Язык: Английский

---

Using natural language processing to analyse text data in behavioural science

Nature Reviews Psychology, Год журнала: 2025, Номер unknown

Опубликована: Янв. 2, 2025

Язык: Английский

---

Blueprint for a Universal Theory of Learning to Read: The Combinatorial Model

Reading Research Quarterly, Год журнала: 2025, Номер 60(2)

Опубликована: Фев. 3, 2025

Abstract In this essay, I outline some of the essential ingredients a universal theory reading acquisition, one that seeks to highlight commonalities while embracing global diversity languages, writing systems, and cultures. begin by stressing need consider insights from multiple disciplines including neurobiology, cognitive science, linguistics, socio‐cultural, historical inquiry, although my major emphasis is on systems approach. A theme common several these perspectives attain level word speed effortlessness necessary overcome severe limitations human (sequential) information processing thereby allowing reader devote maximum resources comprehension. then present Combinatorial Model —a learning read based fundamental principle spoken written language combinatoriality. This (“ infinite ends finite means” ) makes it possible for children learn how decipher (i.e., decode), combine chunk/unitize limited learnable set rudimentary (typically meaningless) elements such as letters, aksharas, syllabograms, character components into nested hierarchy meaningful higher‐order units morphemes words can be recognized instantly effortlessly via rapid parallel their constituent elements. Combinatoriality enables an orthography provide learnability decipherability novice (via phonological transparency well unitizability automatizability expert morphemic ). elaborate (i) dual nature model unfamiliar‐to‐familiar/novice‐to‐expert framework, (ii) unit/s unitization, (iii) writing. liken development tree grows both upwards outwards. Vertical growth thought 3‐phase progression sub‐morphemic, through morpho‐lexical, supra‐lexical phases in which later‐developing do not replace earlier but are added combinatorial hierarchy. Outward conceptualized process knowledge arborization —ongoing refinement, elaboration, diversification. conclude noting that, despite important recent advances, our non‐European non‐alphabetic still its infancy. Current research over‐reliant English—an outlier orthography—together with handful Roman‐script Western European languages. has led science neglect many issues significance homography, tone, diacritics, visual complexity, non‐linearity, linguistic distance, multilingualism, multiscriptism, more. An appreciation specifics particular (or languages) orthographies) child within broader context linguistic, orthographic, cultural crucial only deeper understanding specific truly non‐ethnocentric reading.

Язык: Английский

---