Owing to its exploratory character, the present study was conducted on a sample of students. It was aimed at mapping French students’ views regarding government databases. It used the same theoretical and methodological framework – Information Integration Theory (Anderson, 2008, 2016, 2018) —than the study by Guedez and Mullet (2014). A set of realistic scenarios depicting a situation in which a government planned to create a new, more complete database on people, was created by varying the levels of diverse factors likely to have an impact of people’s views. The first factor considered in the scenarios was the current situation in the country (secure vs. insecure); that is, the government’s probable aims to create a more operational database. It was introduced because promotors of databases’ arguments usually center on security issues: fighting against terrorism, easing the work of the police and the justice system, detecting false identity papers, fighting against identity usurpation (Manenti, 2016).

The second factor considered in the scenarios was the type of information gathered, whether it was only demographic information or complete information involving biometric markers, telephone calls, or and travel habits. This factor is deemed to be an important one because the French National Commission on Informatics and Liberty (CNIL) has been specifically instituted in order to monitor the content of databases and prohibit the introduction of information that could be used to impair the right to privacy and civil liberties. The CNIL’s missions are, among other things: (a) to verify the exactitude of the information, (b) to limit the aims of data gathering and to check whether the nature and amount of data collected match these aims, and (c) to limit the duration of conservation of these data to the attainment of these aims (CNIL, 2018).

The third factor considered was the possibility of access to personal information. According to the Law called “Informatics and Freedom” (Legifrance, 2012), all citizens have the right to access the data contained in their file and to obtain rectification or suppression of these data if they are erroneous. Finally, the fourth factor considered was about the level of confidentiality of the data; that is, whether they have been coded and only public authorities can have access to them or other bodies such as trade companies can take and use them. One of the CNIL’s missions is to guaranty full confidentiality of the data (CNIL, 2018).

Our hypotheses were based on findings from previous studies in empirical ethics that examined the tension between the autonomy principle and the other principles. Three qualitatively different positions were expected that correspond to the three contrasted positions found in these previous studies. For some people, government databases would have no utility and would impede human rights, and human development and survival. This position has been, for example, represented by the group called Pieces et Main-d’Oeuvre (Piece and Manpower) that went as far as requiring the suppression of identity cards and national census (PMO, 2013a; 2013b). Here, the autonomy principle was given priority over the other principles although arguments hostile to the creation and maintenance of databases have been presented as arguments favoring the people’s well-being and the reduction of injustice.

For other people, databases would absolutely be needed as governance tools in a world that is each time more complex. In some cases, these people went as far as requiring the suppression of the CNIL (Barbaux, 2013). Here the benevolence principle has been given priority, without much regards for the autonomy principle. Finally, and probably for most people, the database would be needed to the extent that the information they contain would match the legitimate aims that have motivated their creation. In other words, when information would be limited to these aims, easily accessible to citizens, and fully confidential, databases would be considered as acceptable but when information would not apparently be limited to these aims, not easily accessible to citizens and, as a result, not easily rectifiable, and entirely confidentiality would not be guaranteed, databases would be considered as unacceptable. Here both principles –autonomy and benevolence– would be balanced.

Our second hypothesis was that political orientation would at least partly determine people’s positions regarding the acceptability of government databases. Right-wing people would attribute more importance to the benevolence principle and, more often than others, have a favorable position toward databases. Left-wing people would attribute more importance to the autonomy principle and, more often than others, have an unfavorable position towards databases.

A cluster analysis using the procedure recommended by Hofmans and Mullet (2013) was applied to the raw data to capture possibly different attitudes. As three qualitatively different positions were expected, a three-cluster solution was first tested. Then several other solutions were tested – a two-cluster one, a four-cluster one, a five-cluster one, and a six-cluster one – and findings were compared. The three-cluster solution was discarded, and a four-cluster solution was retained because it was the one that produced the most significant findings. A three-cluster solution could also have been retained, but this solution merged in a single cluster, two groups of participants who had different personal positions.

The main results corresponding to the four clusters are shown in Figure 1, with low acceptability ratings pooled across informative content and possibility of access. Four separate analyses of variance were conducted on the raw data of each cluster with a design of Security x Information x Access x Confidentiality, 2 x 3 x 2 x 3. Owing to the great number of comparisons conducted, the significance threshold was set at 0.001. The main results are shown in Appendix A. Table 1 also shows the composition of the five clusters concerning participants’ gender, year of studies, the type of studies and political orientation.

The first cluster (N = 60, 30% of the sample) was the expected Never acceptable cluster. The overall mean was very low (M = 1.76). Despite that all acceptability ratings were low, ratings were significantly higher when (a) terrorists attacks have taken place (M = 1.89) that when the country was peaceful (M = 1.63), (b) when the information contained in the databases was limited (M = 1.96) than when it was extensive (M = 1.31), and (c) when full confidentiality was guaranteed (M = 2.22) than when it was not (M = 1.30). Besides, the effect of the information factor was stronger when confidentiality was guaranteed (2.60 – 1.95 = 0.65) than when it was not (1.33 – 1.27 = 0.04). Post-licence students and students whose political orientation was Left, or Extreme Left were more often members of this cluster than the license of pre-license students or students whose political orientation was Right.

The second cluster (N = 51, 25% of the sample) was called Tolerable in some cases. The overall mean was low (M = 3.05). Acceptability ratings were significantly higher when (a) terrorists attacks have taken place (M = 3.42) that when the country was peaceful (M = 2.67), (b) when the information contained in the databases was limited (M = 3.53) than when it was extensive (M = 2.57), (c) when people have access to the content of the file (M = 3.35) than when they have no access (M = 2.75), and (d) when full confidentiality was guaranteed (M = 4.08) than when it was not (M = 2.02). In addition, the effect of the information factor was stronger when confidentiality was guaranteed (4.86 – 3.46 = 1.40) than when it was not (2.20 – 1.89 = 0.31), and the effect of the access factor was stronger when confidentiality was guaranteed (4.59 – 3.57 = 1.02) than when it was not (2.11 – 1.92 = 0.19). Mean ratings were higher than the middle of the acceptability scale in only one case: when attacks had taken place, information was limited, and access and confidentiality were guaranteed (M = 6.27). Students whose political orientation was Extreme Left were more often members of this cluster than students whose political orientation was Center.

Figure 1 Figure 1 The four clusters

The third cluster (N = 80, 40% of the sample) was the expected Depends on Guaranties cluster. The overall mean was higher than in both previous clusters (M = 4.68). Acceptability ratings were significantly higher when (a) terrorists attacks have taken place (M = 5.34) that when the country was peaceful (M = 4.01), (b) when the information contained in the databases was limited (M = 5.12) than when it was extensive (M = 4.30), (c) when people have access to the content of the file (M = 5.01) than when they have no access (M = 4.34), and (d) when full confidentiality was guaranteed (M = 6.43) than when it was not (M = 2.92). In addition, the effect of the information factor was stronger when confidentiality was guaranteed (7.15 – 5.83 = 1.40) than when it was not (3.09 – 2.76 = 0.35). Mean ratings were higher than the middle of the acceptability scale in ten cases. Students whose political orientation was Extreme Left were less often members of this cluster than all other students.

The fourth cluster (N = 9, 5% of the sample) was called Undetermined. The overall mean was very close to the middle of the response scale (M = 5.38). No effect was significant. Male students were more often members of this cluster than female students.

An ANOVA was also performed on all the raw data. The design was Political Orientation x Security x Information x Access x Confidentiality, 4 x 2 x 3 x 2 x 2. The results are shown in Table 2. Rightists’ ratings were higher (M = 4.31) than Centrists’ (M = 3.49), Leftists’ (M = 3.31), and Extreme-Leftists’ (M = 2.51) ratings. No interaction involving this factor, however, was significant. The four within-subject effects were significant as well as the three interactions involving the Confidentiality factor (see above).

As expected, a cluster of students expressing a totally negative attitude to government databases was found. Even when all guarantees were present –full confidentiality and free access– and the information collected were only demographic, these students, mostly post-graduate or Left-wingers considered that the creation of government databases was unacceptable. In other words, even the innocuous SSN database would have to be abandoned. This means that the position advocated by groups such as PMO, as extreme as it can appear, is not an isolated position: 30% of the students in our sample shared it.

As also expected, a Depends on Guaranties position was found. For the students sharing this position, mostly Right-wingers and Centrists, government database is acceptable to the condition that confidentiality is secured. Forty percent of students expressed this position, which is consistent with findings in most previous studies on empirical ethics conducted in France that show that people usually take circumstances into account before judging.

An intermediate, not expected position was found that was called Tolerable in some cases. This position was like the Depends on Guaranties position except that the acceptability ratings were located in the center of the acceptability scale; that is, in a zone that could be considered as a tolerability zone. For the students sharing this position, mostly Extreme-Left wingers, the creation of databases would only be, at best, a necessary evil.

Finally, the Always acceptable position was not found. No one student in the sample expressed the view that the creation of databases would be fully liberalized.

As expected, political orientation had a significant effect on students’ positions. Left-wingers were most often of the idea that government databases were either unacceptable or at best tolerable, and Right-wingers and Centrists were most often of the idea that government databases could be acceptable to the essential condition that full confidentiality is guaranteed. Interestingly, political orientation only impacted on the overall attitude to databases; it did not impact on the importance attributed to their characteristics or the circumstances of their creation. Of the four factors considered in the present study, the most important was confidentiality (three-fourths of the explained variance) followed by security (11%), political orientation (6%), type of information (5%) and access (3%), and the order of importance of the four descriptive factors did not vary as a function of political orientation.