Final assignment for the “Data Analysis” course in the Public and Corporate Communication Master’s program (COM), a.y. 2025-2026, Università degli studi di Milano. Instructor: Professor Giulia Maria Dotti Sani.
Introduction
In this paper I use data from the Special Eurobarometer 557 (2025) to investigate social and economic factors influencing attitudes toward Artificial Intelligence (AI) in two European countries with different digital ecosystems: Italy and Denmark.
Literature review
Public attitudes toward AI vary significantly across Europe, reflecting both individual and contextual factors. At the country level, research shows that more digitally advanced ecosystems are associated with higher AI adoption, with Northern European countries leading the way and Southern countries falling behind (Manta et al., 2025; Guarascio et al., 2025). Denmark and Italy are therefore ideal contrasting cases for this comparative analysis.
At the individual level, educational attainment and economic status have been identified among the key drivers of AI attitude, with more educated individuals showing greater openness toward new technologies (Campino & Asvestopoulou, 2025), while those facing financial difficulties may perceive AI as a greater economic threat. Occupation also matters: while AI is generating new professional opportunities (Chen, 2026), many workers perceive it as a threat to their jobs (Bousquette, 2023), particularly those in roles at higher risk of substitution (Ferri et al., 2026; Vu & Lim, 2022).
RQ and Hypothesis
The main research question is: “What social and economic factors are associated with a positive attitude toward AI, and do these associations differ between Italy and Denmark?”
To address it, the following hypotheses are formulated:
H1: People in Denmark are expected to have a more positive attitude toward AI than people in Italy, reflecting Denmark’s more advanced digital ecosystem.
H2: People with a higher level of education have a more positive attitude toward AI than less educated people.
H3: Workers at higher risk of AI substitution have more negative attitudes toward AI than those not exposed to it or those who can benefit from it.
H4: People with fewer financial difficulties have more positive attitudes toward AI than people with greater financial difficulties.
My results confirm that Denmark shows, on average, more positive attitudes toward AI than Italy, and that educational attainment and economic status are significantly associated with attitudes towards AI in both countries. The results for occupation, however, do not meet initial expectations.
Data and methods
For this analysis, I rely on the Special Eurobarometer 557 (September – October 2024), which provides insights into European citizens’ attitudes towards science and technology. The dependent variable asks respondents whether AI will have a positive, negative, or no effect on their way of life in the next 20 years, providing a measure of people’s attitudes toward AI.
The analysis is restricted to adults aged 18 and above in Italy and Denmark. Educational level is captured through a categorical variable distinguishing low, medium, and highly educated respondents. Respondents still studying, refusing to answer, or answering ‘don’t know’ are treated as missing values.
Type of occupation is recoded into three categories based on their relationship with AI: unexposed, at risk of substitution, and empowered by AI (Ferri et al., 2026). Due to the dataset’s lack of detail, however, this classification relies on broad occupational groups, representing a limitation. Those without a paid job are treated as missing values.
As the dataset does not include a variable on income, difficulties in paying bills is used as a proxy for economic status. Due to the limited number of observations in the ‘Most of the time’ category in Denmark (n<30), the variable is recoded into two categories: some difficulties and no difficulties.
Summary statistics and sample sizes are reported in Table 1.
To test all four hypotheses, I use cross-tabulations and chi-square tests, one for each country.
| Italy | Denmark | Total | |
| AI Attitude (%) | |||
| Very negative | 11.77 | 7.08 | 9.44 |
| Fairly negative | 23.33 | 21.24 | 22.29 |
| No effect | 2.33 | 2.90 | 2.61 |
| Fairly positive | 52.70 | 53.11 | 52.91 |
| Very positive | 9.86 | 15.67 | 12.75 |
| N | 943 | 932 | 1.875 |
| Education (%) | |||
| Low | 22.38 | 6.02 | 13.93 |
| Medium | 53.69 | 9.14 | 30.69 |
| High | 23.93 | 84.84 | 55.38 |
| N | 840 | 897 | 1.737 |
| Occupation (%) | |||
| Unexposed | 43.41 | 29.11 | 36.71 |
| At risk | 30.40 | 35.55 | 32.81 |
| Empowered | 26.19 | 35.34 | 30.48 |
| N | 546 | 481 | 1.027 |
| Financial Difficulties (%) | |||
| Some difficulties | 50.53 | 6.76 | 28.69 |
| No difficulties | 49.47 | 93.24 | 71.31 |
| N | 936 | 932 | 1.868 |
Results
Figure 1 shows attitudes toward AI in Italy and Denmark. As expected, a higher proportion of Danes think that AI will have a very positive effect on their lives (0.16 vs 0.10 in Italy), while Italians are more likely to hold very negative views (0.12 vs. 0.07 in Denmark). The chi-square test is significant (X2 = 24.8; p-value<0.001), confirming that AI attitudes differ significantly between the two countries and supporting H1.
Figure 2 shows that in both countries, low educated respondents have more negative attitudes toward AI, accounting for half or more of their group: 0.55 in Italy and 0.50 in Denmark. Medium and highly educated people show more positive attitudes, with similar proportions across countries. Both chi-square tests are significant (Italy: X2 = 72.0, p-value<0.001; Denmark: X2 = 21.4, p-value=0.006), supporting H2. As per Table 1, however, Italy and Denmark show very different education profiles – around 85% of Danes are highly educated vs 24% of Italians – which may partly explain the more positive AI attitudes observed in Denmark (H1).
Figure 3 shows that AI attitudes vary depending on how AI may affect workers’ jobs. In Italy, the chi-square test is significant (X2=27.8, p=0.001), revealing differences across occupational groups. Contrary to H3, however, it is the Unexposed group that holds the most negative attitudes (0.40), rather than those at risk of substitution (0.28), while Empowered workers are the most positive (0.73). In Denmark, differences across groups are smaller and the chi-square test is not significant (X2=9.2, p=0.327). Overall, H3 is not supported in either country, a result that may partly reflect the broad occupational categories used in this analysis.
Figure 4 shows that respondents with no financial difficulties have more positive attitudes toward AI in both countries (Italy 0.66, Denmark 0.70) than those with some difficulties (Italy 0.59; Denmark 0.63). Both chi-square tests are significant (Italy: X2=13.3, p-value=0.010; Denmark: X2=12.6, p-value=0.014), supporting H4. Financial status also differs significantly between the two countries, with only 6% of Danes reporting difficulties in paying bills compared to around 50% of Italians. This may further contribute to explaining the more positive AI attitudes observed in Denmark (H1).
Conclusions
This short paper investigated differences in attitudes towards AI in Italy and in Denmark, analysing both national and individual factors. In Denmark, people have more positive attitudes toward AI than in Italy, supporting H1 and previous studies showing that more digitally advanced ecosystems foster greater openness toward new technologies. The association between educational level and AI attitude is supported in both countries (H2): highly educated people have more positive attitudes than less educated ones. The association between type of occupation and AI attitude is significant only in Italy but does not support H3: contrary to expectations, workers in jobs not exposed to AI hold the most negative attitudes. Finally, the association between financial status and AI attitude is supported in both countries (H4): people with financial difficulties have more negative attitudes toward AI.
The results for H2 and H4 suggest that individual factors play a more determinant role than contextual ones. However, the lack of detail in the occupational variable prevents a precise classification of jobs by their actual AI exposure (Ferri et al., 2026), representing a key limitation of this analysis and suggesting that further research is needed to better understand the role of occupation in shaping AI attitudes across Europe.
References
Bousquette, I. (2023). AI promised to make jobs easier. Workers weren’t so sure. The Wall Street Journal, 27 June.
Campino, J. & Asvestopoulou, E. (2025). AI revolution: trust and the perceived threat to job security. SN Business & Economics, 5(199).
Chen, T. P. (2026). The new jobs being created by AI. The Wall Street Journal, 2 April.
Ferri, V., Porcelli, R. & Fenoaltea, E. M. (2026). Work and artificial intelligence in Italy: between opportunities and risks. International Journal of Manpower, 47(5), pp. 792–831.
Guarascio, D., Reljic, J. & Stöllinger, R. (2025). Diverging paths: AI exposure and employment across European regions. Structural Change and Economic Dynamics, 73, pp. 11–24.
Manta, E. M., Geambasu, M. C. & Birlan, I. (2025). Mapping AI adoption across Europe: a cluster analysis of national responsibility. Proceedings of the International Conference on Business Excellence, 19(1), pp. 1532–1545.
Vu, H. T. & Lim, J. (2022). Effects of country and individual factors on public acceptance of artificial intelligence and robotics technologies: a multilevel SEM analysis of 28-country survey data. Behaviour & Information Technology, 41(7), pp. 1515–1528.
Photo by and machines on Unsplash