Fattorini, E., Rubin, A., Saracino, B., & Bucchi, M. (2026). Italians’ attitudes towards AI: how technology issues travel across social conversation. Information, Communication & Society, 1–23. https://doi.org/10.1080/1369118x.2026.2689030
Summary
This paper analyzes Italian public attitudes toward artificial intelligence using two nationally representative surveys (2023, 2024) from the Observa Science in Society Monitor, conducted during the early diffusion phase of generative AI. The authors argue that Italians exhibit “critical ambivalence” rather than technophobia: they recognize AI’s utility while expressing significant concerns about its risks, with acceptance varying sharply by application domain. Drawing on Bucchi and Trench’s “social conversation around science” framework, the paper shows that AI as a public issue simultaneously occupies multiple communication configurations—dissemination, dialogue, and participation—for different social groups stratified by age, gender, education, and direct experience.
Key Contributions
- Empirical mapping of Italian AI perceptions across two consecutive years, capturing the rapid uptake of generative AI tools.
- Extension of the “social conversation around science” model to AI, demonstrating that a single issue can occupy several communication configurations at once.
- Evidence that AI attitudes are socially stratified rather than uniformly distributed, challenging deficit-model assumptions.
- Policy implications favoring context-sensitive, differentiated communication strategies over uniform information transmission.
- Situates Italy within broader comparative European and international research on AI attitudes.
Methods
Two nationally representative surveys of Italians aged 15+ (May 2023, n=1,011; April 2024, n=1,000), collected via mixed-mode CATI/CAWI with quotas on gender, age, and region. The 2024 wave expanded to 9 questions and 26 items on AI. Analysis combines univariate description, bivariate breakdowns by gender/age/education, and binomial logistic regressions predicting mental imagery, attitudes, views on generative text systems, and perceived long-term societal impacts from demographics, self-perceived knowledge, and use of four AI-related technologies.
Findings
- Self-reported information remains low: only ~3% feel highly informed in 2024, though those feeling “not at all” informed dropped from 15.8% to 4.9%.
- AI tool use grew rapidly: chatbots 11.1%→25.1%, voice assistants 41.3%→49.0%, smart home devices 21.2%→29.0%.
- TV/radio (70.5%) dominates as an AI information source, followed by social media (54.2%) and press (40.0%).
- The humanoid robot remains the most common mental image of AI (27.4%), but chatbot associations grew sharply (15.6%→25.8%).
- Perceiving AI as a threat to humanity rose from 35.8% to 51.0%; belief that machines will replace human activities rose from 33.3% to 46.6%.
- Acceptance is highest for security (76.5%) and lowest for worker management (25.6%), journalism (35.7%), and artistic production (39.6%).
- 69% support permitting generative text systems under strict regulation; only 16.6% favor banning them.
- Regression highlights: university education halves the odds of associating AI with humanoid robots (OR=0.498); chatbot users are far more likely to picture AI as chatbots (OR=3.546); the 45–59 group most fears AI (OR=2.405); women expect fewer positive (OR=0.536) and more negative (OR=1.821) societal effects; chatbot use reduces threat perceptions.
- Only 20.1% expect mostly positive long-term societal effects; 51% expect mixed outcomes.
Connections
No related papers have been provided under shared topics, so there are no genuine intellectual connections to link here. The paper would naturally sit alongside work extending Bucchi and Trench’s science communication framework and comparative survey research on public attitudes toward AI (e.g., Eurobarometer, Pew), but none of those are available as wikilink targets.
Podcast
A research-radio episode discusses this paper: 🎧 MP3 · Spotify · Apple Podcasts