metadata

license: mit
language:
  - en
size_categories:
  - 1K<n<10K

The AI Gap: How Socioeconomic Status Affects Language Technology Interactions

🏆 Best Social Impact Paper Award at ACL 2025

Dataset Summary

This dataset comprises responses from 1,000 individuals from diverse socioeconomic backgrounds, collected to study how socioeconomic status (SES) influences interaction with language technologies, particularly generative AI and large language models (LLMs). Participants shared demographic and socioeconomic data, as well as up to 10 real prompts they previously submitted to LLMs like ChatGPT, totaling 6,482 unique prompts.

Dataset Structure

The dataset is provided as a single CSV file:

survey_language_technologies.csv

Column Name	Description
`id`	Anonymized respondent ID
`gender`	Gender identity (Male, Female, Non-binary, Other, Prefer not to say)
`gender_other`	Custom gender identity if "Other" was selected
`age`	Age group (e.g., 18–24, 25–34, etc.)
`nationality`	One or more nationalities (semicolon-separated)
`ethnicity`	One or more ethnic identities (semicolon-separated)
`ethnicity_other`	Custom ethnicity if "Other" was selected
`marital`	Marital status
`marital_other`	Custom marital status if "Other" was selected
`language`	First language(s) (semicolon-separated)
`language_other`	Custom language if "Other" was selected
`religion`	Religious affiliation
`religion_other`	Custom religion if "Other" was selected
`education`	Participant’s highest education level
`mum_education`	Mother's highest education level
`dad_education`	Father's highest education level
`ses`	Self-assessed SES on a 1–10 ladder scale
`home`	Home ownership status (Own, Rent, Other)
`home_other`	Custom home ownership type if "Other" was selected
`employment`	Current employment status
`occupation`	Current or most recent job (semicolon-separated if multiple)
`mother_occupation`	Mother's occupation(s)
`father_occupation`	Father's occupation(s)
`hobbies`	Hobbies and free-time activities (semicolon-separated)
`hobbies_other`	Custom hobbies if "Other" was selected
`tech`	Daily-used digital devices (semicolon-separated)
`tech_other`	Custom digital devices if "Other" was selected
`know_nlp`	NLP tools the user is familiar with (semicolon-separated)
`know_nlp_other`	Custom tools if "Other" was selected
`use_nlp`	NLP tools the user has used (semicolon-separated)
`use_nlp_other`	Custom tools if "Other" was selected
`would_nlp`	NLP tools find useful but not used because of scance performance (semicolon-separated)
`would_nlp_other`	Custom tools if "Other" was selected
`frequency_llm`	Frequency of LLM use (Every day, Nearly every day, Sometimes, Rarely, Never)
`llm_use`	LLMs used (e.g., ChatGPT, Claude, Bard, etc.)
`llm_other`	Custom LLMs if "Other" was selected
`usecases`	Tasks performed with LLMs (e.g., Writing, Learning, Coding, etc.)
`usecases_other`	Custom tasks if "Other" was selected
`contexts`	Contexts in which LLMs are used (e.g., Work, Personal, School)
`contexts_other`	Custom context if "Other" was selected
`prompt1`–`prompt10`	Up to 10 prompts submitted by the participant to any AI chatbot
`comments`	Open-ended user comments

Note: All multi-select fields are semicolon (;) separated.

Citation

If you use this dataset in your research, please cite the associated paper:

@inproceedings{bassignana-etal-2025-ai,
    title = "The {AI} Gap: How Socioeconomic Status Affects Language Technology Interactions",
    author = "Bassignana, Elisa  and
      Curry, Amanda Cercas  and
      Hovy, Dirk",
    editor = "Che, Wanxiang  and
      Nabende, Joyce  and
      Shutova, Ekaterina  and
      Pilehvar, Mohammad Taher",
    booktitle = "Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
    month = jul,
    year = "2025",
    address = "Vienna, Austria",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2025.acl-long.914/",
    pages = "18647--18664",
    ISBN = "979-8-89176-251-0",
    abstract = "Socioeconomic status (SES) fundamentally influences how people interact with each other and, more recently, with digital technologies like large language models (LLMs). While previous research has highlighted the interaction between SES and language technology, it was limited by reliance on proxy metrics and synthetic data. We survey 1,000 individuals from `diverse socioeconomic backgrounds' about their use of language technologies and generative AI, and collect 6,482 prompts from their previous interactions with LLMs. We find systematic differences across SES groups in language technology usage (i.e., frequency, performed tasks), interaction styles, and topics. Higher SES entail a higher level of abstraction, convey requests more concisely, and topics like `inclusivity' and `travel'. Lower SES correlates with higher anthropomorphization of LLMs (using ``hello'' and ``thank you'') and more concrete language. Our findings suggest that while generative language technologies are becoming more accessible to everyone, socioeconomic linguistic differences still stratify their use to create a digital divide. These differences underscore the importance of considering SES in developing language technologies to accommodate varying linguistic needs rooted in socioeconomic factors and limit the AI Gap across SES groups."
}

Dataset Curators

Elisa Bassignana (IT University of Copenhagen)
Amanda Cercas Curry (CENTAI Institute)
Dirk Hovy (Bocconi University)

Datasets:

MilaNLProc
/

survey-language-technologies

The AI Gap: How Socioeconomic Status Affects Language Technology Interactions

Dataset Summary

Dataset Structure

Citation

Dataset Curators

Links