Changing your mind about the data: Updating sampling assumptions in inductive inference.

When people use samples of evidence to make inferences, they consider both the sample contents and how the sample was generated ("sampling assumptions"). The current studies examined whether people can update their sampling assumptions - whether they can revise a belief about sample generation that is discovered to be incorrect, and reinterpret old data in light of the new belief. We used a property induction task where learners saw a sample of instances that shared a novel property and then inferred whether it generalized to other items. Assumptions about how the sample was selected were manipulated between conditions: in the property sampling frame condition, items were selected because they shared a property, while in the category sampling frame condition, items were selected because they belonged to a particular category. Experiment 1 found that these frames affected patterns of property generalization regardless of whether they were presented before or after the sample data was observed: in both cases, generalization was narrower under a property than a category frame. In Experiments 2 and 3, an initial category or property frame was presented before the sample, and was later retracted and replaced with the complementary frame. Learners were able to update their beliefs about sample generation, basing their property generalization on the more recent correct frame. These results show that learners can revise incorrect beliefs about data selection and adjust their inductive inferences accordingly.

Unifying theories of reasoning and decision making.

De Neys offers a welcome departure from the dual-process accounts that have dominated theorizing about reasoning. However, we see little justification for retaining the distinction between intuition and deliberation. Instead, reasoning can be treated as a case of multiple-cue decision making. Reasoning phenomena can then be explained by decision-making models that supply the processing details missing from De Neys's framework.

What do our sampling assumptions affect: How we encode data or how we reason from it?

In describing how people generalize from observed samples of data to novel cases, theories of inductive inference have emphasized the learner's reliance on the contents of the sample. More recently, a growing body of literature suggests that different assumptions about how a data sample was generated can lead the learner to draw qualitatively distinct inferences on the basis of the same observations. Yet, relatively little is known about how and when these two sources of evidence are combined. Do sampling assumptions affect how the sample contents are encoded, or is any influence exerted only at the point of retrieval when a decision is to be made? We report two experiments aimed at exploring this issue. By systematically varying both the sampling cover story and whether it is given before or after the training stimuli we are able to determine whether encoding or retrieval issues drive the impact of sampling assumptions. We find that the sampling cover story affects generalization when it is presented before the training stimuli, but not after, which suggests that sampling assumptions are integrated during encoding. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

Are logical intuitions only make-believe? Reexamining the logic-liking effect.

An ongoing debate in the literature on human reasoning concerns whether or not the logical status (valid vs. invalid) of an argument can be intuitively detected. The finding that conclusions of logically valid inferences are liked more compared to conclusions of logically invalid ones-called the logic-liking effect-is one of the most prominent pieces of evidence in support of this notion. Trippas et al. (2016) found this logic-liking effect for different kinds of inferences, including conditional and categorical syllogisms. However, all invalid conclusions presented by Trippas et al. (2016) were also impossible given the premises and had a particular structure of surface features-that is, an incongruent atmosphere. We present new data from five preregistered experiments in which we replicate the effect reported by Trippas et al. (2016) for conditional and categorical syllogisms but show that this effect is eliminated when controlling for confounds in surface features. Moreover, we present evidence that there is a demand effect at play, which suggests that people are deliberately considering atmosphere cues of an argument to inform their liking ratings. Taken together, the findings of the present study cast doubt on the existence of logical intuitions. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

Who is sensitive to selection biases in inductive reasoning?

The samples of evidence we use to make inferences in everyday and formal settings are often subject to selection biases. Two property induction experiments examined group and individual sensitivity to one type of selection bias: sampling frames - causal constraints that only allow certain types of instances to be sampled. Group data from both experiments indicated that people were sensitive to the effects of such frames, showing narrower generalization when sample instances were selected because they shared a target property (property sampling) than when instances were sampled because they belonged to a particular group (category sampling). Group generalization patterns conformed to the predictions of a Bayesian model of property induction that incorporates a selective sampling mechanism. In each experiment, however, there was considerable individual variation, with a nontrivial minority showing little sensitivity to sampling frames. Experiment 2 examined correlates of frames sensitivity. A composite measure of working memory capacity predicted individual sensitivity to sampling frames. These results have important implications for current debates about people's ability to factor sample selection mechanisms into their inferences and for the development of formal models of inductive inference. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

Generalization following symmetrical intradimensional discrimination training.

A challenge for generalization models is to specify how excitation generated from a CS+ (i.e., positive evidence) should interact with inhibition from a CS- (i.e., negative evidence) to produce generalized responding. Empirically, many generalization phenomena are consistent with the monotonicity principle, which states that additional positive evidence should increase generalized responding, whereas additional negative evidence should decrease responding. However, a recent study (Lee et al.,, 2019) demonstrated that additional negative evidence can sometimes increase generalization, in direct contrast to animal data and associative accounts of generalization. The current study investigated whether a similar effect could be found in a symmetrical intradimensional discrimination procedure with two sources of negative evidence (CS-s) located on each side of a CS+. In three experiments, we compared generalization along a green-blue dimension between one group of participants who learned that an aqua-colored shape (CS+) predicted an outcome (Single Positive group) with another group who also learned that both a slightly greener and a slightly bluer shape led to no outcome (Double Negative group). Experiments 1A and 1B showed no effect of the additional negative evidence in increasing generalization around the CS+. However, changing a stimulus feature at test (shape) resulted in a higher gradient peak in the Double Negative group relative to the Single Positive group in Experiment 2. Although this result violates the monotonicity principle, an extended version of Blough's (1975) model applying cue competition to multiple stimulus dimensions (i.e., shape and color) successfully replicated the group differences. Our results suggest that associative mechanisms can account for some instances in which negative evidence increases generalization. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

Getting to the source of the illusion of consensus.

Consensus between informants is a valuable cue to a claim's epistemic value, when informants' beliefs are developed independently of each other. Recent work (Yousif et al., 2019) described an illusion of consensus such that people did not generally discriminate between the epistemic warrant of true consensus, where a majority claim is supported by multiple independent sources, and false consensus arising from repetition of a single source's claim. Four experiments tested a novel account of the illusion of consensus; that it arises when people are unsure about the independence of the primary sources on which informant claims are based. When this independence relationship was ambiguous we found evidence for the illusion. However, when steps were taken to highlight the independence between data sources in the true consensus conditions, and confidence in a claim was measured against a no consensus baseline (where there was an equal number of reports supporting and opposing a claim), more weight was given to claims based on true consensus than false consensus. These findings show that although the illusion of consensus is prevalent, people do have the capacity to distinguish between true and false consensus.

Always look on the bright side of logic? Testing explanations of intuitive sensitivity to logic in perceptual tasks.

Much recent research and theorizing in the field of reasoning has been concerned with intuitive sensitivity to logical validity, such as the logic-brightness effect, in which logically valid arguments are judged to have a "brighter" typeface than invalid arguments. We propose and test a novel signal competition account of this phenomenon. Our account makes two assumptions: (a) as per the demands of the logic-brightness task, people attempt to find a perceptual signal to guide brightness judgments, but (b) when the perceptual signal is hard to discern, they instead attend to cues such as argument validity. Experiment 1 tested this account by manipulating the difficulty of the perceptual contrast. When contrast discrimination was relatively difficult, we replicated the logic-brightness effect. When the discrimination was easy, the effect was eliminated. Experiment 2 manipulated the ambiguity of the perceptual task, comparing discrimination performance when the perceptual contrast was labeled in terms of rating "brightness" or "darkness". When the less ambiguous darkness labeling was used, there was no evidence of a logic-brightness effect. In both experiments, individual sensitivity to the perceptual discrimination was negatively correlated with sensitivity to argument validity. Hierarchical latent mixture modeling revealed distinct individual strategies: responses based on perceptual cues, responses based on validity or guessing. Consistent with the signal competition account, the proportion of those responding to validity increased with perceptual discrimination difficulty or task ambiguity. The results challenge explanations of the logic-brightness effect based on parallel dual-process models of reasoning. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

Risking Everything in Obsessive-Compulsive Disorder: An Analogue Decision-Making Study.

An experiment examined decision-making processes among nonclinical participants with low or high levels of OCD symptomatology (N = 303). To better simulate the decision environments that are most likely to be problematic for clients with OCD, we employed decision tasks that incorporated "black swan" options that have a very low probability but involve substantial loss. When faced with a choice between a safer option that involved no risk of loss or a riskier alternative with a very low probability of substantial loss, most participants chose the safer option regardless of OCD symptom level. However, when faced with choices between options that had similar expected values to the previous choices, but where each option had some low risk of a substantial loss, there was a significant shift towards riskier decisions. These effects were stronger when the task involved a contamination based, health-relevant decision task as compared to one with financial outcomes. The results suggest that both low and high symptom OC participants approach decisions involving risk-free options and decisions involving risky alternatives in qualitatively different ways. There was some evidence that measures of impulsivity were better predictors of the shift to risky decision making than OCD symptomatology.

Explaining the popularity bias in online consumer choice.

When people choose products based on online reviews, they show a "popularity bias," overweighting review sample size relative to rated quality. We propose a novel account of this effect based on a causal attribution process, whereby people often interpret larger samples as a proxy for product quality. To test the account, participants in two experiments were asked to rate their product preference based on online reviews showing mean quality scores and review sample sizes for pairs of products. When no explanation for different sample sizes was supplied, we replicated the popularity bias; the product with the larger sample was chosen, even when quality scores modestly favored the alternative. However, as predicted, when sample size differences were explained by a factor unrelated to quality (e.g., time on the market), the popularity bias was substantially reduced. We discuss the implications for models of choice based on social information. (PsycInfo Database Record (c) 2021 APA, all rights reserved).

Modelling generalisation gradients as augmented Gaussian functions.

Studying generalisation of associative learning requires analysis of response gradients measured over a continuous stimulus dimension. In human studies, there is often a high degree of individual variation in the gradients, making it difficult to draw conclusions about group-level trends with traditional statistical methods. Here, we demonstrate a novel method of analysing generalisation gradients based on hierarchical Bayesian curve-fitting. This method involves fitting an augmented (asymmetrical) Gaussian function to individual gradients and estimating its parameters in a hierarchical Bayesian framework. We show how the posteriors can be used to characterise group differences in generalisation and how classic generalisation phenomena such as peak shift and area shift can be measured and inferred. Estimation of descriptive parameters can provide a detailed and informative way of analysing human generalisation gradients.

Adding Types, But Not Tokens, Affects Property Induction.

The extent to which we generalize a novel property from a sample of familiar instances to novel instances depends on the sample composition. Previous property induction experiments have only used samples consisting of novel types (unique entities). Because real-world evidence samples often contain redundant tokens (repetitions of the same entity), we studied the effects on property induction of adding types and tokens to an observed sample. In Experiments 1-3, we presented participants with a sample of birds or flowers known to have a novel property and probed whether this property generalized to novel items varying in similarity to the initial sample. Increasing the number of novel types (e.g., new birds with the target property) in a sample produced tightening, promoting property generalization to highly similar stimuli but decreasing generalization to less similar stimuli. On the other hand, increasing the number of tokens (e.g., repeated presentations of the same bird with the target property) had little effect on generalization. Experiment 4 showed that repeated tokens are encoded and can benefit recognition, but appear to be given little weight when inferring property generalization. We modified an existing Bayesian model of induction (Navarro, Dry, & Lee, 2012) to account for both the information added by new types and the discounting of information conveyed by tokens.

A test of two processes: The effect of training on deductive and inductive reasoning.

Dual-process theories posit that separate kinds of intuitive (Type 1) and reflective (Type 2) processes contribute to reasoning. Under this view, inductive judgments are more heavily influenced by Type 1 processing, and deductive judgments are more strongly influenced by Type 2 processing. Alternatively, single-process theories propose that both types of judgments are based on a common form of assessment. The competing accounts were respectively instantiated as two-dimensional and one-dimensional signal detection models, and their predictions were tested against specifically targeted novel data using signed difference analysis. In two experiments, participants evaluated valid and invalid arguments, under induction or deduction instructions. Arguments varied in believability and type of conditional argument structure. Additionally, we used logic training to strengthen Type 2 processing in deduction (Experiments 1 & 2) and belief training to strengthen Type 1 processing in induction (Experiment 2). The logic training successfully improved validity-discrimination, and differential effects on induction and deduction judgments were evident in Experiment 2. While such effects are consistent with popular dual-process accounts, crucially, a one-dimensional model successfully accounted for the results. We also demonstrate that the one-dimensional model is psychologically interpretable, with the model parameters varying sensibly across conditions. We argue that single-process accounts have been prematurely discounted, and formal modeling approaches are important for theoretical progress in the reasoning field.

Why is logic so likeable? A single-process account of argument evaluation with logic and liking judgments.

Four experiments examined the claims that people can intuitively assess the logical validity of arguments, and that qualitatively different reasoning processes drive intuitive and explicit validity assessments. In each study participants evaluated arguments varying in validity and believability using either deductive criteria (logic task) or via an intuitive, affective response (liking task). Experiment 1 found that people are sensitive to argument validity on both tasks, with valid arguments receiving higher liking as well as higher deductive ratings than invalid arguments. However, the claim that this effect is driven by logical intuitions was challenged by the finding that sensitivity to validity in both liking and logic tasks was affected in similar ways by manipulations of concurrent memory load (Experiments 1 and 2) and variations in individual working memory capacity (Experiments 3 and 4). In both tasks better discrimination between valid and invalid arguments was found when more working memory resources were available. Formal signal detection models of reasoning were tested against the experimental data using signed difference analysis (Stephens, Dunn, & Hayes, 2018b). A single-process reasoning model which assumes that argument evaluation in both logic and liking tasks involves a single latent dimension for assessing argument strength but different response criteria for each task, was found to be consistent with the data from each experiment (as were some dual-process models). The experimental and modeling results confirm that people are sensitive to argument validity in both explicit logic and affect rating tasks, but that these results can be explained by a single underlying reasoning process. (PsycInfo Database Record (c) 2020 APA, all rights reserved).

Stimulus discriminability and induction as independent components of generalization.

Generalization of learning can arise from 2 distinct sources: failure to discriminate a novel test stimulus from the trained stimulus and active extrapolation from the trained stimulus to the test stimulus despite them being discriminable. We investigated these 2 processes in a predictive learning task by testing stimulus discriminability (identification of the trained stimulus) as well as generalization of learning (outcome expectancy). Generalization gradients were broader for expectancy than for identification, in both single cue and differential (discrimination) designs, implying a substantial extrapolation component for the most dissimilar stimuli. The shapes of the expectancy gradients were strongly determined by the training design (single cue vs. differential) and by the rules inferred by participants (similarity vs. linear). By contrast, the identification gradients were unaffected by the training design or inferred rules and were equivalent for predictive and nonpredictive stimuli. These results indicate that perceptual discriminability plays a substantial role in generalization, but it is largely unaffected by associative learning. Instead, learning appears to impact on generalization via an independent extrapolation component which involves cognitive processes such as inductive reasoning. (PsycInfo Database Record (c) 2020 APA, all rights reserved).

Selective sampling and inductive inference: Drawing inferences based on observed and missing evidence.

We propose and test a Bayesian model of property induction with evidence that has been selectively sampled leading to "censoring" or exclusion of potentially relevant data. A core model prediction is that identical evidence samples can lead to different patterns of inductive inference depending on the censoring mechanisms that cause some instances to be excluded. This prediction was confirmed in four experiments examining property induction following exposure to identical samples that were subject to different sampling frames. Each experiment found narrower generalization of a novel property when the sample instances were selected because they shared a common property (property sampling) than when they were selected because they belonged to the same category (category sampling). In line with model predictions, sampling frame effects were moderated by the addition of explicit negative evidence (Experiment 1), sample size (Experiment 2) and category base rates (Experiments 3-4). These data show that reasoners are sensitive to constraints on the sampling process when making property inferences; they consider both the observed evidence and the reasons why certain types of evidence has not been observed.

Evidential diversity increases generalisation in predictive learning.

In property induction tasks, encountering a diverse range of instances (e.g., hippos and hamsters) with a given property usually increases our willingness to generalise that property to a novel instance, relative to non-diverse evidence (e.g., hippos and rhinos). Although generalisation in property induction and predictive learning tasks share conceptual similarities, it is unknown whether this diversity principle applies to generalisation of a predictive association. We tested this hypothesis in two predictive learning experiments using differential training where one category of stimuli (e.g., fruits) predicted an outcome and another category (e.g., vegetables) predicted no outcome. We compared generalisation between a Non-Diverse group who were presented with non-diverse evidence in both positive (predicted the outcome) and negative (predicted no outcome) categories, and two groups who received the same training as the Non-Diverse group but with a more diverse range of exemplars in the positive (Diverse+ group) or negative (Diverse- group) category. Diversity effects were found for both positive and negative categories, in that learning about a diverse range of exemplars increased generalisation of a predictive association to novel exemplars from that same category. The results suggest that diversity, a key principle describing how we reason inductively, also applies to generalisation in associative learning tasks.

The diversity effect in inductive reasoning depends on sampling assumptions.

A key phenomenon in inductive reasoning is the diversity effect, whereby a novel property is more likely to be generalized when it is shared by an evidence sample composed of diverse instances than a sample composed of similar instances. We outline a Bayesian model and an experimental study that show that the diversity effect depends on the assumption that samples of evidence were selected by a helpful agent (strong sampling). Inductive arguments with premises containing either diverse or nondiverse evidence samples were presented under different sampling conditions, where instructions and filler items indicated that the samples were selected intentionally (strong sampling) or randomly (weak sampling). A robust diversity effect was found under strong sampling, but was attenuated under weak sampling. As predicted by our Bayesian model, the largest effect of sampling was on arguments with nondiverse evidence, where strong sampling led to more restricted generalization than weak sampling. These results show that the characteristics of evidence that are deemed relevant to an inductive reasoning problem depend on beliefs about how the evidence was generated.

Belief bias is response bias: Evidence from a two-step signal detection model.

When asked to determine whether a syllogistic argument is deductively valid, people are influenced by their prior beliefs about the believability of the conclusion. Recently, two competing explanations for this belief bias effect have been proposed, each based on signal detection theory (SDT). Under a response bias explanation, people set more lenient decision criteria for believable than for unbelievable arguments. Under the alternative argument strength explanation, believability affects the reasoning stage of processing an argument, with believable and unbelievable arguments differing in subjective strength for both valid and invalid items. Two experiments tested these accounts by asking participants to make validity judgments for categorical syllogisms and to rate their confidence. Conclusion-believability was manipulated both within group (Experiment 1) and between groups (Experiment 2). A novel two-step version of the signal detection model was fit to receiver operating characteristic (ROC) curves for believable and unbelievable arguments. Model fits confirmed that in both experiments there was a shift in decision criterion but not argument discriminability as a function of argument believability. Crucially, when believability is manipulated between groups, this shift is expected under the response bias account but not under the argument strength account. Therefore, the results support the view that belief bias primarily reflects changes in response bias: people require less evidence to endorse a syllogism as valid when it has a believable conclusion. This has important implications for theories of deductive reasoning. (PsycINFO Database Record (c) 2019 APA, all rights reserved).

Negative evidence and inductive reasoning in generalization of associative learning.

When generalizing properties from known to novel instances, both positive evidence (instances known to possess a property) and negative evidence (instances known not to possess a property) must be integrated. The current study compared generalization based on positive evidence alone against a mixture of positive evidence and perceptually dissimilar negative evidence in an interdimensional discrimination procedure. In 2 experiments, we compared generalization following training with a single positive stimulus (that predicted shock) against groups where an additional negative stimulus (that did not predict shock) was presented in a causal judgment (Experiment 1) and a fear conditioning (Experiment 2) procedure. In contrast to animal conditioning studies, we found that adding a "distant" negative stimulus resulted in an overall increase in generalization to stimuli varying on the dimension of the positive stimulus, consistent with the inductive reasoning literature. We show that this key qualitative result can be simulated by a Bayesian model that incorporates helpful sampling assumptions. Our results suggest that similar processes underlie generalization in inductive reasoning and associative learning tasks. (PsycINFO Database Record (c) 2019 APA, all rights reserved).