Thèse Développement de la Vaast en Tant que Mesure Indirecte. H/F - Doctorat.Gouv.Fr

Établissement : Université Grenoble Alpes École doctorale : ISCE - Ingénierie pour la Santé la Cognition et l'Environnement Laboratoire de recherche : Laboratoire interuniversitaire de Psychologie/Personnalité, Cognition, Changement social Direction de la thèse : Dominique MULLER ORCID 0000000195445317 Début de la thèse : 2026-10-01 Date limite de candidature : 2026-05-19T23:59:59 L'une des préoccupations les plus importantes de la psychologie est de prédire le comportement. C'est précisément l'ambition de la notion d'attitude (le plus souvent une évaluation positive ou négative), mesurée pendant longtemps à l'aide de questionnaires, ce qui soulève des problèmes de désirabilité sociale concernant les réponses fournies. Pour pallier cette difficulté, les chercheurs ont développé des mesures implicites ou indirectes, le plus souvent informatisées. Toutefois, bien que ces mesures soient généralement formulées en termes positifs/négatifs, la réaction la plus fondamentale des êtres vivants pourrait plutôt s'exprimer en termes d'approche/évitement. Dans ce projet, nous nous appuierons sur la valeur heuristique d'un modèle de la mémoire issu du cadre de la cognition ancrée afin de proposer une mesure indirecte de ces derniers. During the last few decades, general positive vs. negative inclinations toward stimuli (as social groups) are often assessed with indirect measures. The advantage of these measures lies in the fact that they are supposed to be less influenced by self-presentation concerns compared to direct measures (e.g., self-report). Beyond classic indirect tasks (e.g., IAT affective priming), research on indirect measures of attitudes is constantly expanding through the development of new measurement paradigms. Among them, measures focusing on different aspects of attitudes, such as approach/avoidance tendencies, could constitute an interesting alternative to measure intergroup attitudes. However, if approach/avoidance tasks are to win acclaim as relevant measures of intergroup attitudes, they have to successfully address several central criticisms often expressed toward classic indirect measures (e.g., Gawronski, LeBel, & Peters, 2007). These criticisms are related to the stability of those measures, whether they merely capture cultural knowledge, whether participants are or are not aware of what these measures assess, and finally whether those measures are process pure. Previous research has already applied approach/avoidance measures with social groups. For instance, by using a modified keyboard, Paladino and Castelli (2008) showed several approach/avoidance intergroup effects, with participants-only members of the dominant group (i.e., groups associated with greater social value, such as Whites in most Western societies)-being faster in the compatible setting (i.e., approach ingroup stimuli and avoid outgroup stimuli by moving one's hand toward vs. away from the screen to push keyboard buttons) as compared to the incompatible setting where the approach/avoidance instructions were reversed (see also Castelli, Zogmaister, Smith, & Arcuri, 2004; Clow & Olson, 2010; Degner, Essien, & Reichardt, 2016; Neumann, Hülsenbeck, & Seibt, 2004; Vaes, Paladino, Castelli, Leyens, & Giovanazzi, 2003). So far, however, these approach/avoidance measures have not often been explicitly used as attitude measures. In a preliminary work, we tested in Marine Rougier PhD project several aspects of our approach/avoidance measure-the Visual Approach/Avoidance by the Self Task (VAAST; Rougier et al., 2018)-in the context of social groups (Rougier et al., 2020). We did this, however, with an explicit version of the task where participants had to respond regarding of ingroup vs outgroup first names. Since then, in the PhD project of Yoann Juillard, we developed a non-evaluative version of the VAAST (Julliard et al., 2026). This is the version that we would use in the PhD project.
1. Stability of approach/avoidance compatibility effects assessed with the VAAST
Indirect measures are supposed to capture highly stable evaluation of social groups (Gawronski et al., 2007). Indeed, because indirect measures are supposed to tap more or less automatic evaluations, these automatic evaluations need to be based on years of stored cognitions. First, let us mention that the very idea of stored attitudes or evaluations has been criticized in itself (Schwarz, 2007; Turner, Oakes, Haslam, & McGarty, 1994), notably on the ground of its malleability. Note that a grounded cognition framework like ours can easily agree with the idea that there are no stored fixed attitudes precisely because one of its goals is to defend the idea that knowledge (and also possibly evaluative knowledge) does not have to be stored as representations. Still, something akin to stable attitudes could simply be the results of the fact that the average of memory traces could only change slowly, but it also makes easy to understand that depending on the context different kinds of memory traces can be reactivated. Without inducing such specific contexts, however, one can assume that the average memory traces could be relatively stable. To test this idea with the VAAST, we intend to use a longitudinal design where participants will perform a French/North African first names version twice, with a one-week delay.
2. Cultural knowledge
According to some authors, classic indirect measures would merely capture the dominant cultural message prevalent in a given society, but not the inter-individual variability that may exist among individuals regarding a given attitude object (Arkes & Tetlock, 2004; Olson & Fazio, 2004; Payne, Vuletich, & Lundberg, 2017). Arkes and Tetlock (2004) even argued that If I am aware of the cultural stereotype, I have all the cognitive software that I need to manifest prejudice on the IAT (p. 262). Following this extreme position, it may be argued that classic indirect measures are unable to discriminate individuals who are truly prejudiced (i.e., who personally endorse prevalent cultural knowledge) from individuals who are simply aware of this knowledge by being exposed to the same cultural environment, but do not endorse it personally (Arkes & Tetlock, 2004; see also Gawronski & Bodenhausen, 2006, 2007, 2014; Strack & Deutsch, 2004). Relatedly, in a recent debate, Payne et al. (2017) proposed that classic indirect measures (i.e., the IAT and the AMP) would be more effective in measuring situational biases (e.g., cultural bias) than individual biases (Payne et al., 2017). Two kinds of results could be seen as supporting this point of view. First, among Whites, effects produced by classic indirect measures generally show a surprisingly high (Correll, Park, Judd, & Wittenbrink, 2002; Fazio, Jackson, Dunton, & Williams, 1995; Greenwald et al., 1998; Nosek, Banaji, & Greenwald, 2002) and homogeneous (Judd, Westfall, & Kenny, 2012) bias in favor of Whites. Second, previous work has often found that such pro-White bias is also present among individuals belonging to non-White (i.e., non-dominant) social groups (Ashburn-Nardo, Knowles, & Monteith, 2003; Nosek et al., 2002; Richeson, Trawalter, & Shelton, 2005). Using the Race IAT, Nosek et al. (2002) notably showed, through a very large sample, that Blacks present on average a pro-White bias. These results are consistent with the idea that the social value granted to social groups in a given social context has a large influence on the effects produced by indirect measures of attitudes (Jost, Pelham, & Carvallo, 2002; Rudman, Feinberg, & Fairchild, 2002). Interestingly, in one of the studies we mentioned earlier conducted a French/North-African VAAST with participants from French and North-African origins and this study revealed that both groups showed compatibility effects in the direction of a positive ingroup bias (Rougier et al., 2020). This means that although those participants live within the same cultural context, they showed a compatibility effect favoring their group of origin, unlike it often happens with the IAT.
3. Consciousness of the compatibility effect
Indirect measures have been assumed to provide access to unconscious biases that people may have against their outgroups (Greenwald & Banaji, 1995). Among the evidence against this view may be the most compelling one have been provided by Hahn, Judd, Hirsh, and Blair (2014). These authors first explained their participants how the IAT look like before taking the test and told them that they would perform this test with several groups. Before doing so, however, they were asked to predict the ranking of these groups in terms of their IAT scores. In other words, these authors wanted to test whether participants were in some way conscious of their biases. In contrast with the idea of unconscious biases, the results showed that participants were able to guess those rankings. Here, we will conduct the same kinds of studies with the VAAST to test whether participants can access their biases with the VAAST as they can with the IAT. It is worth mentioning that an additional advantage of this task being that it requires far less trials than a task like the IAT.
4. Decomposing the processes captured by the VAAST
In recent years, scholars have started to criticize the very idea that one can expect to capture one kind of process with direct measures (i.e., more or less controlled processes with paper and pencil measures) and another with indirect measures (i.e., more or less automatic processes with computerized tasks; see for instance, Hütter & Klauer, 2016). Indeed, it very well be that indirect measures in fact capture both types of processes. For instance, when applying this idea to the IAT, Conrey, Sherman, Gawronski, Hugenberg, and Groom (2005) have shown, by relying on multinomial process trees (Hütter & Klauer, 2016), that IAT scores are in fact a compound of four distinct processes. We will apply the same type of general procedure to the VAAST, which will also involve validation studies to validate our interpretation of the different processes.
Overall, if these studies turn out to be promising, this task has already several advantages. Indeed, it is easy to use for researchers, easy to use for participants (it has been used with children), it does not require special material (no specific screen and no specific response devices), it can be used online (again with a user-friendly interface; see Aubé et al., 2019; Julliard et al., 2026), and most importantly it provides highly reliable compatibility effect. The goal is to test the validity of an indirect measure of approach/avoidance automatic action tendencies applied to social stimuli.

Nous recherchons un ou une candidate ayant: de bonnes connaissances en cognition sociale et dans le domain de la méthodologie de la psychologie expérimentale, de bonnes connaissances en traitement des données (si possible avec des notions concernant les modèles mixtes) des compétences en programmation.