Workshop - Rationality, Decision and Epistemic Feelings

Entrée libre

VENDREDI 11 MARS(Salle des Actes, 45 rue d’Ulm)

Matin

Chair : Frédérique de Vignemont
9h30-9h40: Welcome - Introduction

9h40- 10h55: Joëlle Proust (Institut Jean Nicod), 
"Evidence for a dissociation between procedural and belief-based metacognition"

10h55-11h15: Pause Café

11h15-12h30: Paul Egré (Institut Jean Nicod) et Denis Bonnay (Université de Nanterre), 
"On some paradoxes of higher-order knowledge" 

Après-midi
Chair: Tim Bayne

14h15-15h30: Karim N'Diaye & Luc Mallet (Team - Behavior, Emotion, and Basal Ganglia, UPMC - Inserm UMR_S 975 - CNRS UMR 7225, ICM), 
"Pathological doubt and lack of confidence: insights from experimental psychology"

15h30-15h50: Pause Café

15h50-17h05: Sacha Bourgeois-Gironde (Université de Provence & Institut Jan Nicod) et Pierre Grialou (Institut Jean Nicod), 
"Non-conscious Confidence "

17h05-18h20: Jérôme Dokic et Elisabeth Pacherie (Institut Jean Nicod) : 
"Mediating between System 1 and System 2: the role of metacognitive feelings"

SAMEDI 12 MARS (Salle Paul Langevin, 29 rue d’Ulm)

Matin
Chair : Tiziana Zalla

9h40-10h55 : Pascal Mamassian (CNRS et Université Paris Descartes), 
"Uncertainty and confidence in visual perception"

10h55-11h15 Pause Café

11h15-12h30 : Jérôme Sackur (LSCP, ENS), 
Why should we quantify introspection and how should we do it?

ABSTRACTS:

Joëlle Proust (Institut Jean Nicod): 
"Evidence for a dissociation between procedural and belief-based metacognition"

There is a tension, within the domain of metacognitive studies, between two views of the relations between metacognition and mindreading. In a self-ascriptive view, metacognition requires an ability to attribute mental states to oneself. In a self-evaluative view, a large class of metacognitive operations can be developed without metarepresenting one’s mental states. Affective states — sentiments and emotional responses (feelings of doubt, or of conviction) -might provide unarticulated standards of evaluation and, nevertheless, appropriately motivate our epistemic decisions. Evidence for a dissociation between decisions respectively based on epistemic feelings and background beliefs about cognitive ability will be discussed, as it offers a better grasp of the respective roles, in metacognition, of self-evaluation and of self-interpretation. 

Paul Egré (Institut Jean Nicod) & Denis Bonnay (Université de Nanterre) : 
"On some paradoxes of higher-order knowledge"

The aim of this paper is to compare a family of epistemic paradoxes concerning higher-order knowledge, namely sentences of the form "x knows that x knows that p". These paradoxes can all be seen as epistemic analogues of the sorites paradox. The sorites paradox rests on the premise that if a property P applies to some object x, and x is indiscriminably different from y, then P applies to y. Williamson (1992) objected that tolerance is not sound, giving rise to paradox, and advocated replacing it with a weaker principle, the margin for error principle. The margin principle states that if some object x is known to be P, and y is indiscriminably different from x, then y is P. An even weaker version of this principle, discussed by Wright (1992), Fara (2002), and called a gap principle, states that if some object x is known to be P, and y is indiscriminably different from x, then y is not known not to be P. As it turns out, both margin principles and gap principles are incompatible with plausible reliability conditions on knowledge iteration. We discuss the prospects for a uniform solution to these similar, albeit distinct, paradoxes. In a nutshell, this solution relies on the idea that safety conditions for knowledge should be allowed to vary depending on the order of knowledge.

Karim Karim N'Diaye & Luc Mallet (Team - Behavior, Emotion, and Basal Ganglia, UPMC - Inserm UMR_S 975 - CNRS UMR 7225, ICM), 
"Pathological doubt and lack of confidence: insights from experimental psychology"

Doubting about one’s decision or judgments is a common and healthy introspective attitude that may turn out pathologically excessive in some psychiatric conditions, such as obsessive-compulsive disorder (OCD). Multiple cognitive models of OCD suggest that metacognitive dysfunctions might be central to the disorder. Indeed, clinical data, mainly based on self-report questionnaires, have supported the view that metacognitive beliefs (e.g., regarding the consequences of having specific intrusive thoughts) might be deviant in these patients but empirical investigation of metacognitive performance in OCD remains scant. Yet experimental psychology provides us with measures of metacognitive monitoring (accuracy and bias) that have been related to specific neural substrates and that may also be used to assess MC performance in OCD patients, in particular their lack of confidence in their own cognitive functioning. Another aspect of metacognitive skills, namely metacognitive control, has a direct translation in OCD symptomatology: one of the most common compulsions is indeed “checking behavior” that is interpreted as an ill-founded strategy to restore confidence in the outcome of one’s actions. However, empirical evidence suggests -- paradoxically -- that checking may actually decrease confidence. These considerations lead us to formulate a tentative parallel between hyperactive distrustful metacognitive functioning in OCD and addiction and overt habit behaviors.

Sacha Bourgeois-Gironde (Université de Provence & Institut Jan Nicod) & Pierre Grialou (Institut Jean Nicod), 
"Non-conscious Confidence"

We propose arguments in view of revising a standard portrait of confidence in our decisions. Measures of confidence usually rely on introspection, are elicited through retrospective judgments, and are associated with consciousness. In many studies (e.g. Kunimoto et al. 2001), they are even used to assess awareness. 
However, recent arguments and data raise questions on this generally admitted view. For example, studies on monkeys show that confidence, rather than being retrospective, is processed simultaneously with decision (Kiani & Shadlen 2009). Another point is that there are error detection processes, that may be considered as modulators of confidence, as it is traditionally conceived and measured, which arise out of consciousness (Logan & Crump 2010; Salti 2010]. 
Yet, the relationship between these recent results and the standard approach of confidence is not clearly established. Do these non accessible to introspection, unconscious and prospective rather than retrospective aspects of confidence relate to confidence as it is traditionally conceived and measured, or do they indicate the existence of an alternative form of confidence?
We propose that 1/ besides non conscious error detection processes as modulators of explicit and retrospective confidence, there exists a non conscious form of confidence which accompanies our decisions; and that 2/this alternative form of confidence can be dissociated from the one sketched by the standard portrait. 

Jérôme Dokic & Elisabeth Pacherie (Institut Jean Nicod): 
“Mediating between System 1 and System 2: the role of metacognitive feelings”

Over the last two decades, a number of dual processing accounts have been developed in cognitive and social psychology. All these accounts have in common a distinction between cognitive processes that are fast, automatic, high processing capacity and low effort (Type 1) and cognitive processes that are slow, inferential, controlled, limited processing capacity and high effort (Type 2). There are several different hypotheses on how these two types of processes relate. Do they operate in parallel? Are Type 1 processes the default? Or, as has been recently suggested (Evans 2009; Stanovich 2009), do we need to posit a third type of processes (Type 3) to adjudicate between them? Assuming Type 3 processes play this role, do they necessarily operate at the conscious level or not? Here, we suggest that light can be shed on these questions by considering Type 3 processes as metacognitive processes. Some authors (e.g. Thompson 2009) have already proposed that conscious metacognitive feelings (feelings of confidence, rightness, doubt, uncertainty, etc.) determine the intervention of Type 2 processes. However, it is doubtful whether metacognitive processes have to be conscious. Furthermore, it is important to distinguish between metacognitive monitoring and metacognitive control (Koriat 2006). Here we will argue that metacognitive monitoring does not necessarily give rise to conscious feelings and thus that metacognitive control can operate either below or above the threshold of consciousness.

Pascal Mamassian (CNRS & Université Paris Descartes), 
"Uncertainty and confidence in visual perception"

Visual perception is often seen as an inference problem where uncertainty comes from ambiguities in the world (e.g. the two 3D interpretations of the Necker cube), noise in the world (e.g. identifying a scene behind falling snowflakes), or noise in the visual system (e.g. synaptic noise). Uncertainty not only influences our performance in visual perception tasks (e.g. how good are we to discrimination two orientations) but also our confidence in being able to perform these tasks (how well can we predict our performance). We discuss the similarities and differences between performance and confidence in visual perception.

Jérôme Sackur (LSCP, ENS), 
Why should we quantify introspection and how should we do it?

Recently, introspection has attracted a lot of interest both in cognitive psychology and neuroscience. A host of experimental paradigms, measures and models have been proposed, all targeting our meta-cognitive ability to monitor our internal states and processes. This state of affairs is clearly positive, as introspection is undoubtedly an important and long neglected aspect of human cognition. In this talk, I will present the results I obtained with a new experimental paradigm where introspection is operationalized by having participant estimate their own response times. This technique enables us to reconstruct the subjective unfolding of cognitive processes, in a way that partially matches the objective parsing obtained from traditional analyses of objective response times. However, significant discrepancies between objective and subjective temporal unfolding of processes reveal some limits of introspection. I will then relate this experimental measure with other popular measures of introspection (among which confidence is foremost) with a view to establishing a typology. I will stress the often overlooked distinction between subjective and second order measures, and try to establish a methodological road map for introspection research.