Automation scenarios: citizen attitudes towards automated decision-making in the public sector

Figures & data

Table 1. Sample structure of the survey (%).

		Estonia	Germany	Sweden	Total
Gender	Male	51	49	51	50
	Female	49	52	49	50
Age	18–34	26	39	39	34
	35–49	37	38	29	35
	59–75	37	23	32	31
Total	%	100	100	100	100
	N	1500	2001	1000	4501

Figure 1. Employment services scenario index values. A higher score indicates a more positive attitude towards the scenario.

Figure 2. Social services scenario index values. A higher score indicates a more positive attitude towards the scenario.

Figure 3. Predictive policing scenario index values. A higher score indicates a more positive attitude towards the scenario.

Figure 4. Results of a multiple linear regression conducted to predict index values in the employment services scenario. R2 (Adj. R2)  – DE: .14 (.13), EST: .28 (.28), SE: .23 (.22). * = p ≤ 0.05, ** = p ≤ 0.01, *** = p ≤ 0.001.

Figure 5. Results of a multiple linear regression conducted to predict index values in the employment services scenario. R2 (Adj. R2)  – DE: .25 (.24), EST: .24 (.23), SE: .18 (.17). * = p ≤ 0.05, ** = p ≤ 0.01, *** = p ≤ 0.001

Figure 6. Results of a multiple linear regression conducted to predict index values in the predictive policing scenario. R2 (Adj. R2)  – DE: .15 (.14), EST: .14 (.13), SE: .11 (.10). * = p ≤ 0.05, ** = p ≤ 0.01, *** = p ≤ 0.001.

Table

	Completely agree	Rather agree	Rather disagree	Do not agree at all	Don’t know, difficult to say
1. Sorting job seekers automatically will save on costs	4	3	2	1	88
2. Sorting job seekers automatically will be more objective	4	3	2	1	88
3. Sorting job seekers automatically is biased and unfair	4	3	2	1	88

Table

	Completely agree	Rather agree	Rather disagree	Do not agree at all	Don’t know, difficult to say
1. Risk scoring like this will help prevent harm	4	3	2	1	88
2. Harm prevention will be easier and more efficient	4	3	2	1	88
3. Families that have not done anything wrong will be falsely accused of harming their children	4	3	2	1	88

Table

	Completely agree	Rather agree	Rather disagree	Do not agree at all	Don’t know, difficult to say
1. The use of AI for facial recognition is a good way to prevent crime	4	3	2	1	88
2. People who are identified using AI facial recognition should always be thoroughly investigated	4	3	2	1	88
3. If people are wrongly identified and scrutinised, they should be compensated	4	3	2	1	88