The impact of AI errors in a human-in-the-loop process.

Automated decision-making is becoming increasingly common in the public sector. As a result, political institutions recommend the presence of humans in these decision-making processes as a safeguard against potentially erroneous or biased algorithmic decisions. However, the scientific literature on human-in-the-loop performance is not conclusive about the benefits and risks of such human presence, nor does it clarify which aspects of this human-computer interaction may influence the final decision. In two experiments, we simulate an automated decision-making process in which participants judge multiple defendants in relation to various crimes, and we manipulate the time in which participants receive support from a supposed automated system with Artificial Intelligence (before or after they make their judgments). Our results show that human judgment is affected when participants receive incorrect algorithmic support, particularly when they receive it before providing their own judgment, resulting in reduced accuracy. The data and materials for these experiments are freely available at the Open Science Framework:  https://osf.io/b6p4z/ Experiment 2 was preregistered.

Humans inherit artificial intelligence biases.

Artificial intelligence recommendations are sometimes erroneous and biased. In our research, we hypothesized that people who perform a (simulated) medical diagnostic task assisted by a biased AI system will reproduce the model's bias in their own decisions, even when they move to a context without AI support. In three experiments, participants completed a medical-themed classification task with or without the help of a biased AI system. The biased recommendations by the AI influenced participants' decisions. Moreover, when those participants, assisted by the AI, moved on to perform the task without assistance, they made the same errors as the AI had made during the previous phase. Thus, participants' responses mimicked AI bias even when the AI was no longer making suggestions. These results provide evidence of human inheritance of AI bias.

Scarcity affects cognitive biases: The case of the illusion of causality.

Previous research indicates that economic scarcity affects people's judgments, decisions, and cognition in a variety of contexts, and with various consequences. We hypothesized that scarcity could sometimes reduce cognitive biases. Specifically, it could reduce the causal illusion, a cognitive bias that is at the heart of superstitions and irrational thoughts, and consists of believing that two events are causally connected when they are not. In three experiments, participants played the role of doctors deciding whether to administer a drug to a series of patients. The drug was ineffective, because the percentage of patients recovering was identical regardless of whether they took the drug. We manipulated the budget available to buy the drugs, tough all participants had enough for all their patients. Even so, participants in the scarce group reduced the use of the drug and showed a lower causal illusion than participants in the wealthy group. Experiments 2 and 3 added a phase in which the budget changed. Participants who transitioned from scarcity to wealth exhibited a reduced use of resources and a lower causal illusion, whereas participants transitioning from wealth to scarcity were unaffected by their previous history.

When did it happen? Verbal information about causal relations affects time estimation.

Usually, the closer two events occur, the more likely people infer a causal relationship between them. Recent studies have shown that this relationship between time and causality is bidirectional. Participants also tend to judge events closer in time if they assume that they are causally related. We present six experiments showing causal binding, but unlike other experiments, participants do not emit any motor action, and no physical feedback is given. Rather, all stimuli and causal information are provided verbally. After reading a list of events, participants were asked to estimate the time elapsed between two of them. Those participants who were informed that there was a causal relationship between the two events estimated them as occurring closer to each other. These results support causality- and heuristic-based explanations of temporal binding, as opposed to other explanations such as sensory integration or intentionality of action.

Expensive seems better: The price of a non-effective drug modulates its perceived efficacy.

Previous studies have shown that the price of a given product impacts the perceived quality of such product. This finding was also observed in medical contexts, showing that expensive drugs increase the placebo effect compared to inexpensive ones. However, addressing a drug's efficacy requires making causal inferences between the drug and the healing. These inferences rely on the contingency between these two events, a factor that is difficult to control in the placebo research. The present study aimed to test whether the price of a given drug modulates its perceived efficacy using a proper (though fictitious) non-effective drug, so that not only the objective contingency, but also the probability of the cause and the probability of the effect could be adequately controlled for. We expected higher efficacy judgements for the expensive non-effective drug than for the inexpensive one. To test this hypothesis, 60 volunteers participated in a contingency learning task that was programmed so that 72% of the patients healed regardless of whether they took the drug. Approximately one-half of the participants were told that the drug was expensive, whereas the other half were told that it was inexpensive. As expected, the efficacy judgements of participants who saw the expensive drug were significantly higher than those who saw the inexpensive one. Overall, our results showed that the price of a non-effective drug modulates its perceived efficacy, an effect that seems to be mediated by the estimated number of doses administered. This result parallels findings in the placebo literature but using a laboratory methodology that allows stronger control of the variables, suggesting that the illusory overestimation produced by the more expensive treatments might be on the basis of the greater efficacy of the more expensive placebos.

Assessing Emotion and Sensitivity of AI Artwork.

Artificial Intelligence (AI) is currently present in areas that were, until recently, reserved for humans, such as, for instance, art. However, to the best of our knowledge, there is not much empirical evidence on how people perceive the skills of AI in these domains. In Experiment 1, participants were exposed to AI-generated audiovisual artwork and were asked to evaluate it. We told half of the participants that the artist was a human and we confessed to the other half that it was an AI. Although all of them were exposed to the same artwork, the results showed that people attributed lower sensitivity, lower ability to evoke their emotions, and lower quality to the artwork when they thought the artist was AI as compared to when they believed the artist was human. Experiment 2 reproduced these results and extended them to a slightly different setting, a different piece of (exclusively auditory) artwork, and added some additional measures. The results show that the evaluation of art seems to be modulated, at least in part, by prior stereotypes and biases about the creative skills of AI. The data and materials for these experiments are freely available at the Open Science Framework: https://osf.io/3r7xg/. Experiment 2 was preregistered at AsPredicted: https://aspredicted.org/fh2u2.pdf.

Examining potential gender bias in automated-job alerts in the Spanish market.

Numerous field experiments based on the correspondence testing procedure have documented that gender bias influences personnel selection processes. Nowadays, algorithms and job platforms are used for personnel selection processes because of their supposed neutrality, efficiency, and costs savings. However, previous research has shown that algorithms can exhibit and even amplify gender bias. The present research aimed to explore a possible gender bias in automated-job alerts generated in InfoJobs, a popular job platform in Spain. Based on the correspondence testing procedure, we designed eight matched resumes in which we manipulated the gender of the candidate for two different professional sectors (female-dominated vs. male-dominated) and two different levels of age (24 vs. 38). We examined the 3,438 offers received. No significant differences were observed in the automated-job alerts received by female and male candidates as a function of occupation category, salary, and the number of long-term contracts included in the alerts. However, we found significant differences between the female-dominated and the male-dominated sectors in all the mentioned variables. Some limitations and implications of the study are discussed. The data and materials for this research are available at the Open Science Framework, https://osf.io/kptca/.

The influence of algorithms on political and dating decisions.

Artificial intelligence algorithms are ubiquitous in daily life, and this is motivating the development of some institutional initiatives to ensure trustworthiness in Artificial Intelligence (AI). However, there is not enough research on how these algorithms can influence people's decisions and attitudes. The present research examines whether algorithms can persuade people, explicitly or covertly, on whom to vote and date, or whether, by contrast, people would reject their influence in an attempt to confirm their personal freedom and independence. In four experiments, we found that persuasion was possible and that different styles of persuasion (e.g., explicit, covert) were more effective depending on the decision context (e.g., political and dating). We conclude that it is important to educate people against trusting and following the advice of algorithms blindly. A discussion on who owns and can use the data that makes these algorithms work efficiently is also necessary.

The tendency to stop collecting information is linked to illusions of causality.

Previous research proposed that cognitive biases contribute to produce and maintain the symptoms exhibited by deluded patients. Specifically, the tendency to jump to conclusions (i.e., to stop collecting evidence soon before making a decision) has been claimed to contribute to delusion formation. Additionally, deluded patients show an abnormal understanding of cause-effect relationships, often leading to causal illusions (i.e., the belief that two events are causally connected, when they are not). Both types of bias appear in psychotic disorders, but also in healthy individuals. In two studies, we test the hypothesis that the two biases (jumping to conclusions and causal illusions) appear in the general population and correlate with each other. The rationale is based on current theories of associative learning that explain causal illusions as the result of a learning bias that tends to wear off as additional information is incorporated. We propose that participants with higher tendency to jump to conclusions will stop collecting information sooner in a causal learning study than those participants with lower tendency to jump to conclusions, which means that the former will not reach the learning asymptote, leading to biased judgments. The studies provide evidence in favour that the two biases are correlated but suggest that the proposed mechanism is not responsible for this association.

When Success Is Not Enough: The Symptom Base-Rate Can Influence Judgments of Effectiveness of a Successful Treatment.

Patients' beliefs about the effectiveness of their treatments are key to the success of any intervention. However, since these beliefs are usually formed by sequentially accumulating evidence in the form of the covariation between the treatment use and the symptoms, it is not always easy to detect when a treatment is actually working. In Experiments 1 and 2, we presented participants with a contingency learning task in which a fictitious treatment was actually effective to reduce the symptoms of fictitious patients. However, the base-rate of the symptoms was manipulated so that, for half of participants, the symptoms were very frequent before the treatment, whereas for the rest of participants, the symptoms were less frequently observed. Although the treatment was equally effective in all cases according to the objective contingency between the treatment and healings, the participants' beliefs on the effectiveness of the treatment were influenced by the base-rate of the symptoms, so that those who observed frequent symptoms before the treatment tended to produce lower judgments of effectiveness. Experiment 3 showed that participants were probably basing their judgments on an estimate of effectiveness relative to the symptom base-rate, rather than on contingency in absolute terms. Data, materials, and R scripts to reproduce the figures are publicly available at the Open Science Framework: https://osf.io/emzbj/.

PicPsy: A new bank of 106 photographs and line drawings with written naming norms for Spanish-speaking children and adults.

The use of pictures as experimental stimuli is a frequent practice in psychological and educational research. In addition, picture-naming task allows the study of different cognitive processes such as perception, attention, memory and language. Line drawings have been widely used in research to date but it has begun to be highlighted the need for more ecological stimuli such as photographs. However, normative data of a photographic set has not been published yet for use with children. We present PicPsy, a new standardized bank of photographs and matched line drawing. We collected written picture-naming norms for name agreement, unknown responses, alternative names, familiarity and visual complexity. A total of 118 native Spanish-speaking children in grades 3-4 participated in the study. For comparison purposes, 89 adults were also included in the study. Child and adult performance was highly correlated, but we found significant age group differences in all variables examined except for visual complexity. Researchers and teachers could benefit from using the new standardized bank reported here which is published under public domain license. The data and materials for this research are available at the Open Science Framework, https://osf.io/nyf3t/.

Biased Sampling and Causal Estimation of Health-Related Information: Laboratory-Based Experimental Research.

BACKGROUND: The internet is a relevant source of health-related information. The huge amount of information available on the internet forces users to engage in an active process of information selection. Previous research conducted in the field of experimental psychology showed that information selection itself may promote the development of erroneous beliefs, even if the information collected does not. OBJECTIVE: The aim of this study was to assess the relationship between information searching strategy (ie, which cues are used to guide information retrieval) and causal inferences about health while controlling for the effect of additional information features. METHODS: We adapted a standard laboratory task that has previously been used in research on contingency learning to mimic an information searching situation. Participants (N=193) were asked to gather information to determine whether a fictitious drug caused an allergic reaction. They collected individual pieces of evidence in order to support or reject the causal relationship between the two events by inspecting individual cases in which the drug was or was not used or in which the allergic reaction appeared or not. Thus, one group (cause group, n=105) was allowed to sample information based on the potential cause, whereas a second group (effect group, n=88) was allowed to sample information based on the effect. Although participants could select which medical records they wanted to check-cases in which the medicine was used or not (in the cause group) or cases in which the effect appeared or not (in the effect group)-they all received similar evidence that indicated the absence of a causal link between the drug and the reaction. After observing 40 cases, they estimated the drug-allergic reaction causal relationship. RESULTS: Participants used different strategies for collecting information. In some cases, participants displayed a biased sampling strategy compatible with positive testing, that is, they required a high proportion of evidence in which the drug was administered (in the cause group) or in which the allergic reaction appeared (in the effect group). Biased strategies produced an overrepresentation of certain pieces of evidence at the detriment of the representation of others, which was associated with the accuracy of causal inferences. Thus, how the information was collected (sampling strategy) demonstrated a significant effect on causal inferences (F1,185=32.53, P<.001, η2p=0.15) suggesting that inferences of the causal relationship between events are related to how the information is gathered. CONCLUSIONS: Mistaken beliefs about health may arise from accurate pieces of information partially because of the way in which information is collected. Patient or person autonomy in gathering health information through the internet, for instance, may contribute to the development of false beliefs from accurate pieces of information because search strategies can be biased.

Methodological Factors Involved in the Study of Temporal Binding Using the Open Source Software Labclock Web.

Temporal binding occurs when an action and an outcome that follows it after a short period of time are judged as occurring closer to each other in time than they actually are. This effect has often been studied using Libet's clock methodology. Garaizar et al. (2016) presented Labclock Web, a free HTML5 open source software that allows researchers to conduct temporal binding and other experiments using Libet's clock through the Internet. The purpose of the three experiments presented here was to test how certain methodological modifications in the Labclock Web task could impact the temporal binding effect. In comparison with the original study, we aimed to: (a) reduce the interval between action and outcome in the delayed condition to 100 ms, instead of 500, (b) present the two types of trials, immediate and delayed, in two separate consecutive blocks, instead of intermixed, (c) use a visual, rather than auditory, outcome following the action, and (d) reduce the number of trials. In addition to its potential theoretical implications, the results confirm that Labclock Web is a useful and reliable tool for conducting temporal binding experiments and that it is well suited to measure temporal binding effects in a broad range of situations.

Diseases that resolve spontaneously can increase the belief that ineffective treatments work.

RATIONALE: Self-limited diseases resolve spontaneously without treatment or intervention. From the patient's viewpoint, this means experiencing an improvement of the symptoms with increasing probability over time. Previous studies suggest that the observation of this pattern could foster illusory beliefs of effectiveness, even if the treatment is completely ineffective. Therefore, self-limited diseases could provide an opportunity for pseudotherapies to appear as if they were effective. OBJECTIVE: In three computer-based experiments, we investigate how the beliefs of effectiveness of a pseudotherapy form and change when the disease disappears gradually regardless of the intervention. METHODS: Participants played the role of patients suffering from a fictitious disease, who were being treated with a fictitious medicine. The medicine was completely ineffective, because symptom occurrence was uncorrelated to medicine intake. However, in one of the groups the trials were arranged so that symptoms were less likely to appear at the end of the session, mimicking the experience of a self-limited disease. Except for this difference, both groups received similar information concerning treatment effectiveness. RESULTS: In Experiments 1 and 2, when the disease disappeared progressively during the session, the completely ineffective medicine was judged as more effective than when the same information was presented in a random fashion. Experiment 3 extended this finding to a new situation in which symptom improvement was also observed before the treatment started. CONCLUSIONS: We conclude that self-limited diseases can produce strong overestimations of effectiveness for treatments that actually produce no effect. This has practical implications for preventative and primary health services. The data and materials that support these experiments are freely available at the Open Science Framework (https://osf.io/xt3z9/). https://bit.ly/2FMPrMi.

Examining the influence of picture format on children's naming responses.

Digital photography has facilitated the use of more ecological stimuli than line drawings as experimental stimuli. However, there is lack of evidence regarding the effect of the picture format on children's naming agreement. The present work investigated whether the format of presentation of the pictures (line drawing or photograph) affects naming task performance in children. Two naming task experiments are reported using 106 concepts depicted both as a photograph and as a matched drawing delineated directly from the photograph. Thirty-eight and thirty-four Spanish-speaking children from 8 to 10 years old participated in Experiment 1 and Experiment 2, respectively. We examined name agreement measures (H index, percentage of modal name, and alternative responses) and subjective scales (familiarity and visual complexity). The results revealed a significant main effect of format in all of the variables except for familiarity, indicating better name agreement indices and higher visual complexity values for the photograph format than for the line drawing format. Additionally, line drawings were more likely to produce alternative incorrect names. The implications of these findings for psychoeducational research and practice are discussed.

Learning mechanisms underlying accurate and biased contingency judgments.

Many experiments have shown that humans and other animals can detect contingency between events accurately. This learning is used to make predictions and to infer causal relationships, both of which are critical for survival. Under certain conditions, however, people tend to overestimate a null contingency. We argue that a successful theory of contingency learning should explain both results. The main purpose of the present review is to assess whether cue-outcome associations might provide the common underlying mechanism that would allow us to explain both accurate and biased contingency learning. In addition, we discuss whether associations can also account for causal learning. After providing a brief description on both accurate and biased contingency judgments, we elaborate on the main predictions of associative models and describe some supporting evidence. Then, we discuss a number of findings in the literature that, although conducted with a different purpose and in different areas of research, can also be regarded as supportive of the associative framework. Finally, we discuss some problems with the associative view and discuss some alternative proposals as well as some of the areas of current debate. (PsycINFO Database Record (c) 2019 APA, all rights reserved).

Base-rate expectations modulate the causal illusion.

Previous research revealed that people's judgments of causality between a target cause and an outcome in null contingency settings can be biased by various factors, leading to causal illusions (i.e., incorrectly reporting a causal relationship where there is none). In two experiments, we examined whether this causal illusion is sensitive to prior expectations about base-rates. Thus, we pretrained participants to expect either a high outcome base-rate (Experiment 1) or a low outcome base-rate (Experiment 2). This pretraining was followed by a standard contingency task in which the target cause and the outcome were not contingent with each other (i.e., there was no causal relation between them). Subsequent causal judgments were affected by the pretraining: When the outcome base-rate was expected to be high, the causal illusion was reduced, and the opposite was observed when the outcome base-rate was expected to be low. The results are discussed in the light of several explanatory accounts (associative and computational). A rational account of contingency learning based on the evidential value of information can predict our findings.

Thinking in a Foreign language reduces the causality bias.

The purpose of this research is to investigate the impact of a foreign language on the causality bias (i.e., the illusion that two events are causally related when they are not). We predict that using a foreign language could reduce the illusions of causality. A total of 36 native English speakers participated in Experiment 1, 80 native Spanish speakers in Experiment 2. They performed a standard contingency learning task, which can be used to detect causal illusions. Participants who performed the task in their native tongue replicated the illusion of causality effect, whereas those performing the task in their foreign language were more accurate in detecting that the two events were causally unrelated. Our results suggest that presenting the information in a foreign language could be used as a strategy to debias individuals against causal illusions, thereby facilitating more accurate judgements and decisions in non-contingent situations. They also contribute to the debate on the nature and underlying mechanisms of the foreign language effect, given that the illusion of causality is rooted in basic associative processes.

Causal Illusions in the Service of Political Attitudes in Spain and the United Kingdom.

The causal illusion is a cognitive bias that results in the perception of causality where there is no supporting evidence. We show that people selectively exhibit the bias, especially in those situations where it favors their current worldview as revealed by their political orientation. In our two experiments (one conducted in Spain and one conducted in the United Kingdom), participants who self-positioned themselves on the ideological left formed the illusion that a left-wing ruling party was more successful in improving city indicators than a right-wing party, while participants on the ideological right tended to show the opposite pattern. In sum, despite the fact that the same information was presented to all participants, people developed the causal illusion bias selectively, providing very different interpretations that aligned with their previous attitudes. This result occurs in situations where participants inspect the relationship between the government's actions and positive outcomes (improving city indicators) but not when the outcomes are negative (worsening city indicators).

A short educational intervention diminishes causal illusions and specific paranormal beliefs in undergraduates.

Cognitive biases such as causal illusions have been related to paranormal and pseudoscientific beliefs and, thus, pose a real threat to the development of adequate critical thinking abilities. We aimed to reduce causal illusions in undergraduates by means of an educational intervention combining training-in-bias and training-in-rules techniques. First, participants directly experienced situations that tend to induce the Barnum effect and the confirmation bias. Thereafter, these effects were explained and examples of their influence over everyday life were provided. Compared to a control group, participants who received the intervention showed diminished causal illusions in a contingency learning task and a decrease in the precognition dimension of a paranormal belief scale. Overall, results suggest that evidence-based educational interventions like the one presented here could be used to significantly improve critical thinking skills in our students.