Answers are grounded in this paper and cite their sources.
Abstract concepts are frequently mapped onto more concrete sensorimotor experiences. One such mapping concerns time and space: people who read and write from left to right tend to associate the past with the left and the future with the right, a correspondence known as the mental timeline (MTL). The present study examined whether an MTL-consistent spatial bias is also observed in ChatGPT and whether this bias can be altered by manipulating linguistic and cultural context. In four experiments, thirty-two independent chat sessions were conducted, with each session treated as an independent AI respondent and asked to place past-related and future-related words at any location on an xy plane (a word allocation task). When the prompt was written in English (Experiment 1), written in Arabic (Experiment 2), or written in Arabic and additionally specified that the model was a native speaker of Arabic (Experiment 3), past words were placed to the left and future words to the right. In contrast, when a Japanese prompt specified that the model was a person from a culture in which text is read from right to left (Experiment 4), this pattern was reversed. These results suggest that ChatGPT predominantly exhibits the MTL typical of English-speaking cultures, and that this mapping can be modulated by framing reading direction.
No takes yet. Share an insight, caveat, or question.
Uchikawa et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: