Harvard finds gaps in AI training, support and everyday use
A report combines surveys, browsing records and news coverage to examine women’s participation in AI in the US. Access to training and workplace support emerges as a central difference.
Summary
Making AI tools available does not give everyone equal opportunities to understand and use them. A Harvard Shorenstein Center report, published on October 14, identifies differences between women and men in workplace support, frequency of use and visibility as experts.
The research combines a survey of 5,403 US adults, browsing records and an analysis of 78,542 news articles. These sources allow researchers to compare self-reported behavior with other evidence of use and participation in public discussion.
In practice
In the survey, 29% of women and 22% of men said they had never used AI. Desktop browsing records, however, showed no statistically significant difference in the proportion who had tried these tools. They did show more intensive use among men who used them. Trying a tool once and incorporating it into everyday activity are different measures.
Among employed respondents, women reported less access to AI training, mentorship and encouragement. These forms of support were associated with greater use and confidence for both groups. For an organization, this shifts attention toward who receives time, guidance and opportunities to learn, alongside who has an account.
The disparity also appears in public authority: men were cited as experts 4.3 times as often. Even after removing people cited mainly because of their position, women remained a minority among sources selected for their knowledge.
Context
The authors do not argue that greater AI use or trust should be goals in themselves. Privacy and the reliability of answers are major concerns for both groups. The issue is whether people have the support to decide when the technology adds value.
The full report and methodology document the samples and differences between measures. The findings describe the US and should not be automatically applied to Portugal. An association between support and use also does not, by itself, establish causation.
Why it matters
- Adoption depends on learning conditions as well as access to tools.
- Counting accounts alone can conceal differences in use and support.
- Selecting experts for news, events and training influences who helps shape AI.