Mental State Inferences Research Articles

How do we think machines think? An fMRI study of alleged competition with an artificial intelligence

Mentalizing is defined as the inference of mental states of fellow humans, and is a particularly important skill for social interactions. Here we assessed whether activity in brain areas involved in mentalizing is specific to the processing of mental states or can be generalized to the inference of non-mental states by comparing brain responses during the interaction with an intentional and an artificial agent. Participants were scanned using fMRI during interactive rock-paper-scissors games while believing their opponent was a fellow human (Intentional agent, Int), a humanoid robot endowed with an artificial intelligence (Artificial agent, Art), or a computer playing randomly (Random agent, Rnd). Participants' subjective reports indicated that they adopted different stances against the three agents. The contrast of brain activity during interaction with the artificial and the random agents didn't yield any cluster at the threshold used, suggesting the absence of a reproducible stance when interacting with an artificial intelligence. We probed response to the artificial agent in regions of interest corresponding to clusters found in the contrast between the intentional and the random agents. In the precuneus involved in working memory, the posterior intraparietal suclus, in the control of attention and the dorsolateral prefrontal cortex, in executive functions, brain activity for Art was larger than for Rnd but lower than for Int, supporting the intrinsically engaging nature of social interactions. A similar pattern in the left premotor cortex and anterior intraparietal sulcus involved in motor resonance suggested that participants simulated human, and to a lesser extend humanoid robot actions, when playing the game. Finally, mentalizing regions, the medial prefrontal cortex and right temporoparietal junction, responded to the human only, supporting the specificity of mentalizing areas for interactions with intentional agents.

The Development of Distrust

Preschool-age children's reasoning about the reliability of deceptive sources was investigated. Ninety 3- to 5-year-olds watched several trials in which an informant gave advice about the location of a hidden sticker. Informants were either helpers who were happy to give correct advice, or trickers who were happy to give incorrect advice. Three-year-olds tended to accept all advice from both helpers and trickers. Four-year-olds were more skeptical but showed no preference for advice from helpers over trickers, even though they differentiated between helpers and trickers on metacognitive measures. Five-year-olds systematically preferred advice from helpers. Selective trust was associated with children's ability to make mental state inferences.

Cultural influences on neural basis of intergroup empathy

Cultures vary in the extent to which people prefer social hierarchical or egalitarian relations between individuals and groups. Here we examined the effect of cultural variation in preference for social hierarchy on the neural basis of intergroup empathy. Using cross-cultural neuroimaging, we measured neural responses while Korean and American participants observed scenes of racial ingroup and outgroup members in emotional pain. Compared to Caucasian-American participants, Korean participants reported experiencing greater empathy and elicited stronger activity in the left temporo-parietal junction (L-TPJ), a region previously associated with mental state inference, for ingroup compared to outgroup members. Furthermore, preferential reactivity within this region to the pain of ingroup relative to outgroup members was associated with greater preference for social hierarchy and ingroup biases in empathy. Together, these results suggest that cultural variation in preference for social hierarchy leads to cultural variation in ingroup-preferences in empathy, due to increased engagement of brain regions associated with representing and inferring the mental states of others.

Identifying the What, Why, and How of an Observed Action: An fMRI Study of Mentalizing and Mechanizing during Action Observation

Humans commonly understand the unobservable mental states of others by observing their actions. Embodied simulation theories suggest that this ability may be based in areas of the fronto-parietal mirror neuron system, yet neuroimaging studies that explicitly investigate the human ability to draw mental state inferences point to the involvement of a “mentalizing" system consisting of regions that do not overlap with the mirror neuron system. For the present study, we developed a novel action identification paradigm that allowed us to explicitly investigate the neural bases of mentalizing observed actions. Across repeated viewings of a set of ecologically valid video clips of ordinary human actions, we manipulated the extent to which participants identified the unobservable mental states of the actor (mentalizing) or the observable mechanics of their behavior (mechanizing). Although areas of the mirror neuron system did show an enhanced response during action identification, its activity was not significantly modulated by the extent to which the observers identified mental states. Instead, several regions of the mentalizing system, including dorsal and ventral aspects of medial pFC, posterior cingulate cortex, and temporal poles, were associated with mentalizing actions, whereas a single region in left lateral occipito-temporal cortex was associated with mechanizing actions. These data suggest that embodied simulation is insufficient to account for the sophisticated mentalizing that human beings are capable of while observing another and that a different system along the cortical midline and in anterior temporal cortex is involved in mentalizing an observed action.

How does your own knowledge influence the perception of another person’s action in the human brain?

When you see someone reach into a cookie jar, their goal remains obvious even if you know that the last cookie has already been eaten. Thus, it is possible to infer the goal of an action even if you know that the goal cannot be achieved. Previous research has identified distinct brain networks for processing information about object locations, actions and mental-state inferences. However, the relationship between brain networks for action understanding in social contexts remains unclear. Using functional magnetic resonance imaging, this study assesses the role of these networks in understanding another person searching for hidden objects. Participants watched movie clips depicting a toy animal hiding and an actor, who was ignorant of the hiding place, searching in the filled or empty location. When the toy animal hid in the same location repeatedly, the blood oxygen level-dependent (BOLD) response was suppressed in occipital, posterior temporal and posterior parietal brain regions, consistent with processing object properties and spatial attention. When the actor searched in the same location repeatedly, the BOLD signal was suppressed in the inferior frontal gyrus, consistent with the observation of hand actions. In contrast, searches towards the filled location compared to the empty location were associated with a greater response in the medial prefrontal cortex and right temporal pole, which are both associated with mental state inference. These findings show that when observing another person search for a hidden object, brain networks for processing information about object properties, actions and mental state inferences work together in a complementary fashion. This supports the hypothesis that brain regions within and beyond the putative human mirror neuron system are involved in action comprehension within social contexts.

Triangles have goals too: Understanding action representation in left aIPS

Humans freely interpret moving shapes as being “alive” and having social intentions, such as beliefs and desires. The brain systems underpinning these processes are the same as those used to detect animacy and infer mental states from human behaviour. However, it is not yet known if the brain systems that respond to human action-goals also respond to the action-goals of shapes. In the present paper, we used a repetition suppression paradigm during functional magnetic resonance imaging (fMRI) to examine brain systems that respond to the action-goals of shapes. Participants watched video clips of simple, geometrical shapes performing different ‘take-object’ goals. Repeated presentation of the same goal suppressed the blood oxygen level-dependent (BOLD) response in left anterior intraparietal sulcus (aIPS), a brain region known to distinguish the goals of human hand actions. This finding shows that left aIPS shows similar sensitivity to the action-goals of human and non-human agents. Our data complement previous work on animacy perception and mental state inference, which suggest components of the social brain are driven by the type of action comprehension that is engaged rather than by the form of the acting agent (i.e., human or shape). Further, the results have consequence for theories of goal understanding in situations without access to biological form or motion.

Broadening the Scope of Affect Detection Research

I propose to broaden the scope of affect detection research in three related directions. First, the task definition should be broadened from the detection of affects to the inference of mental states. Second, the detection process should be reconceptualized as an inference to the best explanation that makes essential use of a theory of mind. Third, additional data sources should be utilized to infer emotions: information about the situation (eliciting events), information about emotion-related goal-directed actions, and self-reports about emotions.

Student mental state inference from unintentional body gestures using dynamic Bayesian networks

Applications that interact with humans would benefit from knowing the intentions or mental states of their users. However, mental state prediction is not only uncertain but also context dependent. In this paper, we present a dynamic Bayesian network model of the temporal evolution of students’ mental states and causal associations between mental states and body gestures in context. Our approach is to convert sensory descriptions of student gestures into semantic descriptions of their mental states in a classroom lecture situation. At model learning time, we use expectation maximization (EM) to estimate model parameters from partly labeled training data, and at run time, we use the junction tree algorithm to infer mental states from body gesture evidence. A maximum a posteriori classifier evaluated with leave-one-out cross validation on labeled data from 11 students obtains a generalization accuracy of 97.4% over cases where the student reported a definite mental state, and 83.2% when we include cases where the student reported no mental state. Experimental results demonstrate the validity of our approach. Future work will explore utilization of the model in real-time intelligent tutoring systems.

Neural correlates of social cognition in naturalistic settings: A model-free analysis approach

Neuroimaging studies have consistently identified a network of brain regions subserving inferences of other humans' mental states. This network consists of the superior temporal sulcus, temporoparietal junction, medial prefrontal cortex, temporal poles, and precuneus. Little is known, however, about the neural substrate underlying Theory of Mind processes in close to real-life conditions. To investigate those processes in more naturalistic settings, we used an fMRI adaption of the video-based Movie for the Assessment of Social Cognition (MASC; Dziobek et al., 2006), which considers separate analysis of implicit mental state reasoning during rapidly changing perceptual cues as demanded in naturalistic settings and explicit mental state reasoning. We analyzed fMRI data by means of both a standard general linear model (GLM) approach and a tensor probabilistic independent component analysis (T-PICA), which is a novel model-free approach that allows decomposition of activation into independent spatio-temporally coherent functional networks. The model-based GLM approach revealed the typical explicit mental state reasoning network. Complementary to the GLM approach, the model-free T-PICA approach showed that those regions are also recruited during implicit mental state reasoning and that they are represented in three independent, functionally connected networks. The first component, mediating face processing and recognition, comprises the occipito-parietotemporal cortices, while the second component, involved in language comprehension, comprises the temporal lobes, lateral prefrontal cortex, and precuneus. The dorsomedial prefrontal cortex and the precuneus comprise the third component, which is likely responsible for self-referential mental activity.These results show that the mental state reasoning network can be decomposed into circumscribed functional networks mediating differential aspects of Theory of Mind.

Action understanding as inverse planning

Humans are adept at inferring the mental states underlying other agents’ actions, such as goals, beliefs, desires, emotions and other thoughts. We propose a computational framework based on Bayesian inverse planning for modeling human action understanding. The framework represents an intuitive theory of intentional agents’ behavior based on the principle of rationality: the expectation that agents will plan approximately rationally to achieve their goals, given their beliefs about the world. The mental states that caused an agent's behavior are inferred by inverting this model of rational planning using Bayesian inference, integrating the likelihood of the observed actions with the prior over mental states. This approach formalizes in precise probabilistic terms the essence of previous qualitative approaches to action understanding based on an “intentional stance” [Dennett, D. C. (1987). The intentional stance. Cambridge, MA: MIT Press] or a “teleological stance” [Gergely, G., Nádasdy, Z., Csibra, G., & Biró, S. (1995). Taking the intentional stance at 12 months of age. Cognition, 56, 165–193]. In three psychophysical experiments using animated stimuli of agents moving in simple mazes, we assess how well different inverse planning models based on different goal priors can predict human goal inferences. The results provide quantitative evidence for an approximately rational inference mechanism in human goal inference within our simplified stimulus paradigm, and for the flexible nature of goal representations that human observers can adopt. We discuss the implications of our experimental results for human action understanding in real-world contexts, and suggest how our framework might be extended to capture other kinds of mental state inferences, such as inferences about beliefs, or inferring whether an entity is an intentional agent.

An fMRI Investigation of Spontaneous Mental State Inference for Moral Judgment

Human moral judgment depends critically on "theory of mind," the capacity to represent the mental states of agents. Recent studies suggest that the right TPJ (RTPJ) and, to lesser extent, the left TPJ (LTPJ), the precuneus (PC), and the medial pFC (MPFC) are robustly recruited when participants read explicit statements of an agent's beliefs and then judge the moral status of the agent's action. Real-world interactions, by contrast, often require social partners to infer each other's mental states. The current study uses fMRI to probe the role of these brain regions in supporting spontaneous mental state inference in the service of moral judgment. Participants read descriptions of a protagonist's action and then either (i) "moral" facts about the action's effect on another person or (ii) "nonmoral" facts about the situation. The RTPJ, PC, and MPFC were recruited selectively for moral over nonmoral facts, suggesting that processing moral stimuli elicits spontaneous mental state inference. In a second experiment, participants read the same scenarios, but explicit statements of belief preceded the facts: Protagonists believed their actions would cause harm or not. The response in the RTPJ, PC, and LTPJ was again higher for moral facts but also distinguished between neutral and negative outcomes. Together, the results illuminate two aspects of theory of mind in moral judgment: (1) spontaneous belief inference and (2) stimulus-driven belief integration.

Mentalizing under Uncertainty: Dissociated Neural Responses to Ambiguous and Unambiguous Mental State Inferences

The ability to read the minds of others (i.e., to mentalize) requires that perceivers understand a wide range of different kinds of mental states, including not only others’ beliefs and knowledge but also their feelings, desires, and preferences. Moreover, although such inferences may occasionally rely on observable features of a situation, perceivers more typically mentalize under conditions of “uncertainty,” in which they must generate plausible hypotheses about a target's mental state from ambiguous or otherwise underspecified information. Here, we use functional neuroimaging to dissociate the neural bases of these 2 distinct social–cognitive challenges: 1) mentalizing about different types of mental states (beliefs vs. preferences) and 2) mentalizing under conditions of varying ambiguity. Although these 2 aspects of mentalizing have typically been confounded in earlier research, we observed a double dissociation between the brain regions sensitive to type of mental state and ambiguity. Whereas ventral and dorsal aspects of medial prefrontal cortex responded more during ambiguous than unambiguous inferences regardless of the type of mental state, the right temporoparietal junction was sensitive to the distinction between beliefs and preferences irrespective of certainty. These results underscore the emerging consensus that, rather than comprising a single mental operation, social cognition makes flexible use of different processes as a function of the particular demands of the social context.

Looking at theory-of-mind 'on-line': A new procedure to track mental state inferences in young children

Event Abstract Back to Event Looking at theory-of-mind 'on-line': A new procedure to track mental state inferences in young children Maria Núñez1* 1 Dept of Psychology, Glasgow Caledonian University, United Kingdom For 25 years, research on theory-of-mind (ToM) has consolidated the fact that young children attribute mental states to themselves and others in order to understand and predict their behaviour. The assessment of these inferential processes, however, is typically made through behavioral measures that test the child’s inferences ‘off-line’ (usually through verbal responses to a critical question). This procedure has proved useful in revealing the early onset of this species-specific ability but it misses out one of the main features of their use in real life, namely, the rapid, ‘on-line’ and often implicit nature of these processes. The series of studies presented here are part of a research project that aims to track mental states inferences on-line and put temporal measures in relation to the usual verbal measures. In two studies, 3 to 5 yr-olds are presented with a set of tasks using a new procedure that tracks their mental-state inferences on-line while children follow a narrative in a computerised, self-administered task. Narratives are accompanied by sets of pictures. The story is “unfolded” in the computer screen as the child clicks the mouse. Card next to the last in the episode-sequence contains information that is either consistent or inconsistent with the inferred ‘mental state’ (either emotions or beliefs, depending on the study and task). Children’s response times (RTs), their pattern of visual inspection (using a TOBII eye-tracker) and their justifications following the consistent/inconsistent episode are assessed. Initial results show that overall: (1) four- and 5- year olds’ RTs are faster than those of 3-year-olds; (2) while 4- and 5-yr-olds ‘move on’ more quickly in Consistent (as compared to Inconsistent) episodes, 3-yr-olds present the opposite pattern. (3) Initial analysis of their visual inspection patterns also indicates differences between younger and older pre-schoolers (4) Verbal justifications refer mostly to the key event under the Consistent condition while they are randomly distributed for the Inconsistent condition. Further analyses are still in progress but, as a whole, findings show a coherent and complementary pattern of responses across the set of measures, revealing the RTs and visual inspection patterns as sensitive measures to approach the assessment of the inferences on-line. Discussion will include how this procedure can be adapted to simultaneously record ERPs in order to really capture the on-line nature of our mental state reasoning and to help us to determine to what extent this relies on conscious processes. Conference: 10th International Conference on Cognitive Neuroscience, Bodrum, Turkey, 1 Sep - 5 Sep, 2008. Presentation Type: Poster Presentation Topic: Perceptual Processing and Recognition Citation: Núñez M (2008). Looking at theory-of-mind 'on-line': A new procedure to track mental state inferences in young children. Conference Abstract: 10th International Conference on Cognitive Neuroscience. doi: 10.3389/conf.neuro.09.2009.01.330 Copyright: The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers. They are made available through the Frontiers publishing platform as a service to conference organizers and presenters. The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated. Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed. For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions. Received: 15 Dec 2008; Published Online: 15 Dec 2008. * Correspondence: Maria Núñez, Dept of Psychology, Glasgow Caledonian University, Glasgow, United Kingdom, m.nunez@gcal.ac.uk Login Required This action requires you to be registered with Frontiers and logged in. To register or login click here. Abstract Info Abstract The Authors in Frontiers Maria Núñez Google Maria Núñez Google Scholar Maria Núñez PubMed Maria Núñez Related Article in Frontiers Google Scholar PubMed Abstract Close Back to top Javascript is disabled. Please enable Javascript in your browser settings in order to see all the content on this page.

Learning to fear what others have feared before

Learning to fear what others have feared before

Reading minds versus following rules: Dissociating theory of mind and executive control in the brain

The false belief task commonly used in the study of theory of mind (ToM) requires participants to select among competing responses and inhibit prepotent responses, giving rise to three possibilities: (1) the false belief tasks might require only executive function abilities and there may be no domain-specific component; (2) executive control might be necessary for the emergence of ToM in development but play no role in adult mental state inferences; and (3) executive control and domain-specific ToM abilities might both be implicated. We used fMRI in healthy adults to dissociate these possibilities. We found that non-overlapping brain regions were implicated selectively in response selection and belief attribution, that belief attribution tasks recruit brain regions associated with response selection as much as well-matched control tasks, and that regions associated with ToM (e.g., the right temporo-parietal junction) were implicated only in the belief attribution tasks. These results suggest that both domain-general and domain-specific cognitive resources are involved in adult ToM.

Response Times of Children and Adolescents with Asperger Syndrome on an ‘Advanced’ Test of Theory of Mind

In the present study the response times of 10- to 20-year-old participants with Asperger syndrome (AS) (N=21) of normal intelligence and a control group of typically developing individuals (N=20) were recorded on a new 'advanced' test of theory of mind. This test taps the ability to make mental-state inferences versus physical-state inferences in a story context. The participants with AS were significantly slower than the controls on both tasks. In addition, the differences in response times between mental- and physical-state inference were significantly larger in the AS group than in the control group, suggesting that the clinical group experienced more problems than the controls in making inferences about mental states than about physical states.

Dissociable Medial Prefrontal Contributions to Judgments of Similar and Dissimilar Others

Human social interaction requires the recognition that other people are governed by the same types of mental states-beliefs, desires, intentions-that guide one's own behavior. We used functional neuroimaging to examine how perceivers make mental state inferences when such self-other overlap can be assumed (when the other is similar to oneself) and when it cannot (when the other is dissimilar from oneself). We observed a double dissociation such that mentalizing about a similar other engaged a region of ventral mPFC linked to self-referential thought, whereas mentalizing about a dissimilar other engaged a more dorsal subregion of mPFC. The overlap between judgments of self and similar others suggests the plausibility of "simulation" accounts of social cognition, which posit that perceivers can use knowledge about themselves to infer the mental states of others.

Mirror neurons and imitation: A computationally guided review

Neurophysiology reveals the properties of individual mirror neurons in the macaque while brain imaging reveals the presence of ‘mirror systems’ (not individual neurons) in the human. Current conceptual models attribute high level functions such as action understanding, imitation, and language to mirror neurons. However, only the first of these three functions is well-developed in monkeys. We thus distinguish current opinions (conceptual models) on mirror neuron function from more detailed computational models. We assess the strengths and weaknesses of current computational models in addressing the data and speculations on mirror neurons (macaque) and mirror systems (human). In particular, our mirror neuron system (MNS), mental state inference (MSI) and modular selection and identification for control (MOSAIC) models are analyzed in more detail. Conceptual models often overlook the computational requirements for posited functions, while too many computational models adopt the erroneous hypothesis that mirror neurons are interchangeable with imitation ability. Our meta-analysis underlines the gap between conceptual and computational models and points out the research effort required from both sides to reduce this gap.

FMRI Study of Emotional Speech Comprehension

Little is known about the neural correlates of affective prosody in the context of affective semantic discourse. We used functional magnetic resonance imaging to investigate this issue while subjects performed 1) affective classification of sentences having an affective semantic content and 2) grammatical classification of sentences with neutral semantic content. Sentences of each type were produced half by actors and half by a text-to-speech software lacking affective prosody. Compared with neutral sentences processing, sentences with affective semantic content--with or without affective prosody--led to an increase in activation of a left inferior frontal area involved in the retrieval of semantic knowledge. In addition, the posterior part of the left superior temporal sulcus (STS) together with the medial prefrontal cortex were recruited, although not activated by neutral sentences classification. Interestingly, these areas have been described as implicated during self-reflection or other's mental state inference that possibly occurred during the affective classification task. When affective prosody was present, additional rightward activations of the human-selective voice area and the posterior part of STS were observed, corresponding to the processing of speaker's voice emotional content. Accurate affective communication, central to social interactions, requires the cooperation of semantics, affective prosody, and mind-reading neural networks.

Mentalizing and Marr: An information processing approach to the study of social cognition

To interact successfully, individuals must not only recognize one another as intentional agents driven primarily by internal mental states, but also possess a system for making reliable and useful inferences about the nature of those beliefs, feelings, goals, and dispositions. The ability to make such mental state inferences (i.e., to mentalize or mindread) is the central accomplishment of human social cognition. The present article suggests that our understanding of how humans go about making mental state inferences will benefit from treating social cognition primarily as an information processing system that comprises a set of mechanisms for elaborating more basic social information into an understanding of another's mind. Following Marr's [Marr, D., 1982. Vision. W. H. Freeman, San Francisco, CA] framework for the study of such information processing systems, I suggest that questions about social cognition might profitably be asked at three levels – computation, algorithm, and implementation – and outline a number of ways in which a description of social cognition at the middle level (i.e., the step-by-step processes that give rise to mental state inferences) can be informed by analysis at the other two.