In many critical investigations of machine vision, the focus lies almost exclusively on dataset bias and on fixing datasets by introducing more and more diverse sets of images. We propose that machine vision systems are inherently biased not only because they rely on biased datasets but also because theirperceptual topology, their specific way of representing the visual world, gives rise to a new class of bias that we callperceptual bias. Concretely, we define perceptual topology as the (...) set of those inductive biases in machine vision systems that determine its capability to represent the visual world. Perceptual bias, then, describes the difference between the assumed “ways of seeing” of a machine vision system, our reasonable expectations regarding its way of representing the visual world, and its actual perceptual topology. We show how perceptual bias affects the interpretability of machine vision systems in particular, by means of a close reading of a visualization technique called “feature visualization”. We conclude that dataset bias and perceptual bias both need to be considered in the critical analysis of machine vision systems and propose to understand critical machine vision as an important transdisciplinary challenge, situated at the interface of computer science and visual studies/Bildwissenschaft. (shrink)

Perceptual learning is commonly understood as conferring some benefit to the learner, such as allowing for the extraction of more information from the environment. However, perceptual learning can be biased in several different ways, some of which do not appear to provide such a benefit. Here we outline a systematic framework for thinking about bias in perceptual learning and discuss how several cases fit into this framework. We argue these biases are compatible with an understanding in (...) which perceptual learning is beneficial, but that its benefits are tied to both a person's narrow interests and the training environment or domain, and so if there are changes to either of these, then benefits can turn into liabilities, though these are often temporary. (shrink)

Siegel recently distinguishes between seven possible ways in which our perceptual access to social information can be biased by flawed practice of either individuals or social structures, two of which, namely attention and cognitive penetration, imply that it is the content of perception, as opposed to that of judgments, that is biased. Both attention and cognitive penetration, however, rely on cognitive states imposing top-down influences on perceptual states. As such, perceptual bias resulting from them is to (...) a large extent merely a derivation of cognitive bias. In this paper, I propose another way in which our perception can be biased, namely, as the result of faulty assumptions made by the visual system. Furthermore, I argue that in contrast to cognitive penetration and attentional direction, perceptual bias arising in this way is fundamentally perceptual and does not depend on inputs from one’s cognitive system. This sort of perceptual bias, if it exists, would have important implications for how we conceptualize social bias and pose special challenges to traditional interventions designed to counteract bias. (shrink)

In the typical single-stimulus perceptual identification task, accuracy is improved by prior study of test words, a repetition priming benefit. There is also a cost, inasmuch as previously studied words are likely to be produced as responses if the test word is orthographically similar but not identical to a studied word. In two-alternative forced-choice perceptual identification, a test word is flashed and followed by two alternatives, one of which is the correct response. When the two alternatives are orthographically (...) similar, test words identical to previously studied items are identified more accurately than new words but tests words orthographically similar to studied words are identified less accurately than new words . Ratcliff and McKoon argue that these are bias effects that arise in the decision stage of word identification. We report five experiments that examined the alternative hypothesis that these bias effects arise from postperceptual guessing strategies. In single-stimulus perceptual identification, repetition priming benefits were equally great for young and older adults who claimed to use deliberate guessing strategies and those who did not . In contrast, only groups of young and older people who claimed to deliberately guess studied words in a two-alternative forced-choice task showed reliable benefits and costs. Costs and benefits were abolished in the two-alternative forced-choice task when a very long study list was used, presumably because the increased retrieval burden made the use of deliberate guessing strategies less attractive . Under conditions similar to those of Experiment 3, repetition priming was observed in single-stimulus perceptual identification . These results are consistent with the view that costs and benefits in the forced-choice perceptual identification task arise from deliberate guessing strategies but that those in the single-stimulus task do not. The possibility that the observed relationship between strategy reports and priming effects reflects erroneous postexperimental assessments of strategies by participants is also considered. (shrink)

chapter on perception and bias including implicit bias.

What is the mental representation that is responsible for implicit bias? What is this representation that mediates between the trigger and the biased behavior? My claim is that this representation is neither a propositional attitude nor a mere association. Rather, it is mental imagery: perceptual processing that is not directly triggered by sensory input. I argue that this view captures the advantages of the two standard accounts without inheriting their disadvantages. Further, this view also explains why manipulating mental (...) imagery is among the most efficient ways of counteracting implicit bias. (shrink)

Over the course of her career, Jean Harvey argued that as agents engaged in a “life of moral endeavor,” we should understand ourselves and others to be moral works in progress, always possessing the potential to grow beyond and become more than the sum of our past wrongs. In this paper I follow Harvey and argue that in order to live a life of moral endeavor, it is not enough merely to know about injustice. Instead, we must engage in the (...) difficult and often painful task of overcoming deep-seated cognitive biases that cause us to fail to perceive the ubiquitous injustice that pervades our world. I conclude by arguing that education, empathy, and love can each help us to increase our perceptual awareness of injustice and so should be recognized to be crucial parts of a life of moral endeavor. (shrink)

There is an active debate within the field of phonology concerning the cognitive status of substantive phonetic factors such as ease of articulation and perceptual distinctiveness. A new framework is proposed in which substance acts as a bias, or prior, on phonological learning. Two experiments tested this framework with a method in which participants are first provided highly impoverished evidence of a new phonological pattern, and then tested on how they extend this pattern to novel contexts and novel (...) sounds. Participants were found to generalize velar palatalization (e.g., the change from [k] as in keep to [t⌢∫S] as in cheap) in a way that accords with linguistic typology, and that is predicted by a cognitive bias in favor of changes that relate perceptually similar sounds. Velar palatalization was extended from the mid front vowel context (i.e., before [e] as in cape) to the high front vowel context (i.e., before [i] as in keep), but not vice versa. The key explanatory notion of perceptual similarity is quantified with a psychological model of categorization, and the substantively biased framework is formalized as a conditional random field. Implications of these results for the debate on substance, theories of phonological generalization, and the formalization of similarity are discussed. (shrink)

Visual perception relies on stored information and environmental associations to arrive at a determinate representation of the world. This opens up the disturbing possibility that our visual experiences could themselves be subject to a kind of racial bias, simply in virtue of accurately encoding previously encountered environmental regularities. This possibility raises the following question: what, if anything, is wrong with beliefs grounded upon these prejudicial experiences? They are consistent with a range of epistemic norms, including evidentialist and reliabilist standards (...) for justification. I argue that we will struggle to locate a flaw with these sorts of perceptual beliefs so long as we focus our analysis at the level of the individual and her response to information. We should instead broaden our analysis to include the social structure within which the individual is located. Doing so lets us identify a problem with the way in which unjust social structures in particular “gerrymander” the regularities an individual is exposed to, and by extension the priors their visual system draws on. I argue that in this way, social structures can cap perceptual skill. (shrink)

In this review I describe research conducted in my laboratory concerning implicit mood-congruent memory (MCM) bias in clinical depression. MCM is the tendency for depressed individuals to retrieve more unpleasant information from memory than nondepressed controls, and may be an important maintenance mechanism in depression. MCM has been studied frequently with explicit memory tests, but relatively few studies have investigated MCM using implicit memory tests. I describe several implicit memory studies which show that: (a) an implicit MCM bias (...) does not appear to exist when perceptually driven tests are used; (b) implicit memory bias can be found when conceptually driven tests are used, but (c) not all conceptually driven tests show implicit MCM bias. I conclude that conceptual processing is necessary, but is not sufficient for demonstrating implicit memory bias in depression. Future studies should investigate specific components of conceptual elaboration that support implicit memory bias in depression. (shrink)

Using the matching bias example, the aim of the present studies was to show that adults' reasoning biases are due to faulty executive inhibition programming. In the first study, the subjects were trained on Wason's classical card selection task; half were given training in how to inhibit the perceptual matching bias (experimental group) and half in logic without the inhibition component (control group). On the pre- and post-tests, their performance was assessed on the Evans conditional rule falsification (...) task (with a negation in the antecedent of the rule), a task that also involves matching bias. In addition, subjects were tested for perceptual field dependence/independence using the Embedded Figures Test. The results brought out a specific inhibition training effect, as well as a clear-cut relationship in the experimental group between receptiveness to training and perceptual field independence. In the second study, the training paradigm was the same except that on the pre- and post-tests, the negation was in the consequent of the conditional rule (in this case, the perceptual matching response corresponds to the logical response). The subjects succeeded on the pre-test, and the matching-bias inhibition training had a negative effect on post-test performance. This specific negative priming effect confirms the inhibitory impact of our experimental training and outlines the dissociation of inhibition and logical components. (shrink)

Autism spectrum disorder is characterized by atypical perception, including processing that is biased toward local details rather than global configurations. This bias may impact on memory. The present study examined the effect of this perception on both implicit and explicit memory in conditions that promote either local or global processing. The first experiment consisted of an object identification priming task using two distinct encoding conditions: one favoring local processing and the other favoring global processing of drawings. The second experiment (...) focused on episodic memory with two different cartoon recognition tasks that favored either local or a global processing. In addition, all the participants underwent a general clinical cognitive assessment aimed at documenting their cognitive profile and enabling correlational analyses with experimental memory tasks. Seventeen participants with ASD and 17 typically developing controls aged from 10 to 16 years participated to the first experiment and 13 ASD matched with 13 TD participants were included for the second experiment. Experiment 1 confirmed the preservation of priming effects in ASD but, unlike the Comparison group, the ASD group did not increase his performance as controls after a globally oriented processing. Experiment 2 revealed that local processing led to difficulties in discriminating lures from targets in a recognition task when both lures and targets shared common details. The correlation analysis revealed that these difficulties were associated with processing speed and inhibition. These preliminary results suggest that natural perceptual processes oriented toward local information in ASD may impact upon their implicit memory by preventing globally oriented processing in time-limited conditions and induce confusion between explicit memories that share common details. (shrink)

Aging is accompanied by stereotyped changes in functional brain activations, for example a cortical shift in activity patterns from posterior to anterior regions is one hallmark revealed by functional magnetic resonance imaging (fMRI) of aging cognition. Whether these neuronal effects of aging could potentially contribute to an amelioration of or resistance to the cognitive symptoms associated with psychopathology remains to be explored. We used a visual illusion paradigm to address whether aging affects the cortical control of perceptual beliefs and (...) biases. Our aim was to understand the effective connectivity associated with volitional control of ambiguous visual stimuli and to test whether greater top-down control of early visual networks emerged with advancing age. Using a bias training paradigm for ambiguous images we found that older participants (n = 16) resisted experimenter-induced visual bias compared to a younger cohort (n = 14) and that this resistance was associated with greater activity in prefrontal and temporal cortices. By applying Dynamic Causal Models for fMRI we uncovered a selective recruitment of top-down connections from the middle temporal to Lingual gyrus (LIN) by the older cohort during the perceptual switch decision following bias training. In contrast, our younger cohort did not exhibit any consistent connectivity effects but instead showed a loss of driving inputs to orbitofrontal sources following training. These findings suggest that perceptual beliefs are more readily controlled by top-down strategies in older adults and introduce age-dependent neural mechanisms that may be important for understanding aberrant belief states associated with psychopathology. (shrink)

Multimodal integration is the formation of a coherent percept from different sensory inputs such as vision, audition, and somatosensation. Most research on multimodal integration in speech perception has focused on audio-visual integration. In recent years, audio-tactile integration has also been investigated, and it has been established that puffs of air applied to the skin and timed with listening tasks shift the perception of voicing by naive listeners. The current study has replicated and extended these findings by testing the effect of (...) air puffs on gradations of voice onset time along a continuum rather than the voiced and voiceless endpoints of the original work. Three continua were tested: bilabial (“pa/ba”), velar (“ka/ga”), and a vowel continuum (“head/hid”) used as a control. The presence of air puffs was found to significantly increase the likelihood of choosing voiceless responses for the two VOT continua but had no effect on choices for the vowel continuum. Analysis of response times revealed that the presence of air puffs lengthened responses for intermediate (ambiguous) stimuli and shortened them for endpoint (non-ambiguous) stimuli. The slowest response times were observed for the intermediate steps for all three continua, but for the bilabial continuum this effect interacted with the presence of air puffs: responses were slower in the presence of air puffs, and faster in their absence. This suggests that during integration auditory and aero-tactile inputs are weighted differently by the perceptual system, with the latter exerting greater influence in those cases where the auditory cues for voicing are ambiguous. (shrink)

Light and Kennison found that bias effects in the forced-choice perceptual identification of words occurred only in a subset of participants, those who claimed on a strategy questionnaire to be deliberately guessing words they had studied previously. McKoon and Ratcliff raised a number of objections to the proposal that bias effects are due to guessing strategies, citing difficulties in our statistical treatment of data, our use of subjective reports to classify participants, and our approach to the general (...) problem of separating implicit from explicit influences on performance. This article responds to these objections. (shrink)

Human decision making shows systematic simplifications and deviations from the tenets of rationality (‘heuristics’) that may lead to suboptimal decisional outcomes (‘cognitive biases’). There are currently three prevailing theoretical perspectives on the origin of heuristics and cognitive biases: a cognitive-psychological, an ecological and an evolutionary perspective. However, these perspectives are mainly descriptive and none of them provides an overall explanatory framework for the underlying mechanisms of cognitive biases. To enhance our understanding of cognitive heuristics and biases we propose a neural (...) network framework for cognitive biases, which explains why our brain systematically tends to default to heuristic (‘Type 1’) decision making. We argue that many cognitive biases arise from intrinsic brain mechanisms that are fundamental for the working of biological neural networks. In order to substantiate our viewpoint, we discern and explain four basic neural network principles: (1) Association, (2) Compatibility (3) Retainment, and (4) Focus. These principles are inherent to (all) neural networks which were originally optimized to perform concrete biological, perceptual, and motor functions. They form the basis for our inclinations to associate and combine (unrelated) information, to prioritize information that is compatible with our present state (such as knowledge, opinions and expectations), to retain given information that sometimes could better be ignored, and to focus on dominant information while ignoring relevant information that is not directly activated. The supposed mechanisms are complementary and not mutually exclusive. For different cognitive biases they may all contribute in varying degrees to distortion of information. The present viewpoint not only complements the earlier three viewpoints, but also provides a unifying and binding framework for many cognitive bias phenomena. (shrink)

Radical skepticism is the view that we know nothing, or at least next to nothing. Nearly no one actually believes that skepticism is true. Yet it has remained a serious topic of discussion for millennia and it looms large in popular culture. What explains its persistent and widespread appeal? How does the skeptic get us to doubt what we ordinarily take ourselves to know? I present evidence from two experiments that classic skeptical arguments gain potency from an interaction between two (...) factors. First, people evaluate inferential belief more harshly than perceptual belief. Second, people evaluate inferential belief more harshly when its content is negative (i.e. that something is not the case) than when it’s positive (i.e. that something is the case). It just so happens that potent skeptical arguments tend to focus our attention on negative inferential beliefs, and we are especially prone to doubt that such beliefs count as knowledge. That is, our cognitive evaluations are biased against this specific combination of source and content. The skeptic sows seeds of doubt by exploiting this feature of our psychology. (shrink)

Radical skepticism is the view that we know nothing or at least next to nothing. Nearly no one actually believes that skepticism is true. Yet it has remained a serious topic of discussion for millennia and it looms large in popular culture. What explains its persistent and widespread appeal? How does the skeptic get us to doubt what we ordinarily take ourselves to know? I present evidence from two experiments that classic skeptical arguments gain potency from an interaction between two (...) factors. First, people evaluate inferential belief more harshly than perceptual belief. Second, people evaluate inferential belief more harshly when its content is negative (i.e., that something is not the case) than when it is positive (i.e., that something is the case). It just so happens that potent skeptical arguments tend to focus our attention on negative inferential beliefs, and we are especially prone to doubt that such beliefs count as knowledge. That is, our cognitive evaluations are biased against this specific combination of source and content. The skeptic sows seeds of doubt by exploiting this feature of our psychology. (shrink)

We report two experiments in which participants had to judge the time of occurrence of a stimulus relative to a clock. The experiments were based on the control condition used by Libet, Gleason, Wright, and Pearl [Libet, B., Gleason, C. A., Wright, E. W., & Pearl, D. K. . Time of conscious intention to act in relation to onset of cerebral activities : The unconscious initiation of a freely voluntary act. Brain 106, 623–642] to correct for any bias in (...) the estimation of the time at which an endogenous event, the conscious intention to perform a movement, occurred. Participants’ responses were affected systematically by the sensory modality of the stimulus and by the speed of the clock. Such findings demonstrate the variability in judging the time at which an exogenous event occurs and, by extension, suggest that such variability may also apply to the judging the time of occurrence of endogenous events. The reliability of participants’ estimations of when they formed the conscious intention to perform a movement in Libet et al.’s study is therefore questionable. (shrink)

We use a transmission chain method to establish how context and category salience influence the formation of novel stereotypes through cumulative cultural evolution. We created novel alien targets by combining features from three category dimensions—color, movement, and shape—thereby creating social targets that were individually unique but that also shared category membership with other aliens (e.g., two aliens might be the same color and shape but move differently). At the start of the transmission chains each alien was randomly assigned attributes that (...) described it (e.g., arrogant, caring, confident). Participants were given training on the alien‐attribute assignments and were then tested on their memory for these. The alien‐attribute assignments participants produced during test were used as the training materials for the next participant in the transmission chain. As information was repeatedly transmitted an increasingly simplified, learnable stereotype‐like structure emerged for targets who shared the same color, such that by the end of the chains targets who shared the same color were more likely to share the same attributes (a reanalysis of data from Martin et al., which we term Experiment 1). The apparent bias toward the formation of novel stereotypes around the color category dimension was also found for objects (Experiment 2). However, when the category dimension of color was made less salient, it no longer dominated the formation of novel stereotypes (Experiment 3). The current findings suggest that context and category salience influence category dimension salience, which in turn influences the cumulative cultural evolution of information. (shrink)

Neurobiological and cognitive models of unconscious information processing suggest that subconscious threat detection can lead to cognitive misinterpretations and false alarms, while conscious processing is assumed to be perceptually and conceptually accurate and unambiguous. Furthermore, clinical theories suggest that pathological anxiety results from a crude preattentive warning system predominating over more sophisticated and controlled modes of processing. We investigated the hypothesis that subconscious detection of threat in a cognitive task is reflected by enhanced ''false signal'' detection rather than by selectively (...) enhanced discrimination of threat items in 30 patients with panic disorder and 30 healthy controls. We presented a tachistoscopic word-nonword discrimination task and a subsequent recognition task and analyzed the data by means of process-dissociation procedures. In line with our expectations, subjects of both groups showed more false signal detection to threat than to neutral stimuli as indicated by an enhanced response bias, whereas indices of discriminative sensitivity did not show this effect. In addition, patients with panic disorder showed a generally enhanced response bias in comparison to healthy controls. They also seemed to have processed the stimuli less elaborately and less differentially. Results are consistent with the assumption that subconscious threat detection can lead to misrepresentations of stimulus significance and that pathological anxiety is characterized by a hyperactive preattentive alarm system that is insufficiently controlled by higher cognitive processes. (shrink)

While the recognition of ambiguous emotions is crucial for successful social interactions, previous work has shown that they are perceived differently depending on whether they are viewed on male or female faces. The present paper aims to shed light on this phenomenon by exploring two hypotheses: the confounded signal hypothesis, which posits the existence of perceptual overlaps between emotions and gendered morphotypes, and the social role hypothesis, according to which the observer's responses are biased by stereotypes. Participants were asked (...) to categorise blended faces (i.e. artificial faces made ambiguous by mixing two emotions) in a forced-choice task. Six emotions were used to create each blend (neutral, surprise, sadness, fear, happiness, anger), for a total of 15 expressions. We then applied signal detection theory – considering both the morphotype of the stimuli and the participants’ gender – to distinguish participants’ perceptual processes from their response biases. The results showed a perceptual advantage for anger on male faces and for sadness on female faces. However, different strategies were deployed when labelling emotions on gendered morphotypes. In particular, a response bias towards angry male faces establishes their special status, as they resulted in both excellent detection and a tendency to be over-reported, especially by women. (shrink)

Light and Kennison proposed that bias effects in the forced-choice perceptual identification of words result from a strategy engaged in by subjects to retrieve explicit information about the words. This article enumerates several problems with this proposal and presents new experimental data against it. It is concluded that subjects do not ordinarily employ an explicit retrieval strategy. The data are discussed in the context of the general problem of separating implicit from explicit influences on performance.

When visual attention is directed away from a stimulus, neural processing is weak and strength and precision of sensory data decreases. From a computational perspective, in such situations observers should give more weight to prior expectations in order to behave optimally during a discrimination task. Here we test a signal detection theoretic model that counter-intuitively predicts subjects will do just the opposite in a discrimination task with two stimuli, one attended and one unattended: when subjects are probed to discriminate the (...) unattended stimulus, they rely less on prior information about the probed stimulus’ identity. The model is in part inspired by recent findings that attention reduces trial-by-trial variability of the neuronal population response and that they use a common criterion for attended and unattended trials. In five different visual discrimination experiments, when attention was directed away from the target stimulus, subjects did not adjust their response bias in reaction to a change in stimulus presentation frequency despite being fully informed and despite the presence of performance feedback and monetary and social incentives. This indicates that subjects did not rely more on the priors under conditions of inattention as would be predicted by a Bayes-optimal observer model. These results inform and constrain future models of Bayesian inference in the human brain. (shrink)

What is the relation between Kanizsa's bias towards convexity and the Gestaltists' demonstrations that perceptual organization obeys a principle of pragnänz, or simplicity? Why should either kind of bias exist? Textbook accounts assign no functional role for these biases. Geon theory (Biederman 1987) proposes that we can understand these biases in terms of fundamental processes by which complex objects are decomposed into convex (or singly concave) regions at points of matched cusps according to the transversality regularity (Hoffman (...) and Richards 1985). Such decomposition yields simple, convex parts segmented between the concavities. A shape that contains concavities is generally regarded as complex insofar as it can be decomposed into the regions, or parts, between the concavities. It is these simple parts that are the stable elements of shape, not the whole object. In fact, geon theory leads to the expectation that shape recognition proceeds most efficiently when the parts are good (in the pragnänz sense) but the object is bad! (shrink)

In higher-level cognitive tasks, older compared to younger adults show a bias towards positive emotion information and away from negative information (a positivity effect). It is unclear whether this effect occurs in early perceptual processing. This issue is important for determining if the positivity effect is due to automatic rather than controlled processing. We tested this with older and younger adults on a positive/negative face emotion valence classification task using masked priming. Positive (happy) and negative (angry) face targets (...) were preceded by masked repetition or valence primes with neutral face baselines. In Experiment 1, 30 younger and 30 older adults were tested with 50 ms primes. Younger adults showed repetition priming for both positive and negative targets. Older adults showed repetition priming for positive but not negative targets. Neither group showed valence priming. In Experiment 2, 30 older and 29 younger adults were tested with longer duration primes. Younger adults showed repetition priming for both positive and negative emotions, and no valence priming. Older adults only showed repetition and valence priming for positive targets. We proposed older adults’ lack of angry face priming was due to an early attention orienting strategy favouring happy expressions at the expense of angry ones. (shrink)

We report two experiments in which participants had to judge the time of occurrence of a stimulus relative to a clock. The experiments were based on the control condition used by Libet, Gleason, Wright, and Pearl [Libet, B., Gleason, C. A., Wright, E. W., & Pearl, D. K. . Time of conscious intention to act in relation to onset of cerebral activities : The unconscious initiation of a freely voluntary act. Brain 106, 623–642] to correct for any bias in (...) the estimation of the time at which an endogenous event, the conscious intention to perform a movement, occurred. Participants’ responses were affected systematically by the sensory modality of the stimulus and by the speed of the clock. Such findings demonstrate the variability in judging the time at which an exogenous event occurs and, by extension, suggest that such variability may also apply to the judging the time of occurrence of endogenous events. The reliability of participants’ estimations of when they formed the conscious intention to perform a movement in Libet et al.’s study is therefore questionable. (shrink)