Interference between cues of the same outcome depends on the causal interpretation of the events

In an interference-between-cues design, the expression of a learned Cue A → Outcome 1 association has been shown to be impaired if another cue, B, is separately paired with the same outcome in a second learning phase. In the present study, we assessed whether this interference effect is mediated by participants' previous causal knowledge. This was achieved by having participants learn in a diagnostic situation in Experiment 1a, and then by manipulating the causal order of the learning task in Experiments 1b and 2. If participants use their previous causal knowledge during the learning process, interference should only be observed in the diagnostic situation because only there we have a common cause (Outcome 1) of two disjoint effects, namely cues A and B. Consistent with this prediction, interference between cues was only found in Experiment 1a and in the diagnostic conditions of Experiments 1b and 2.     

A dissociation between causal judgment and outcome recall

It has been suggested that causal learning in humans is similar to Pavlovian conditioning in animals. According to this view, judgments of cause reflect the degree to which an association exists between the cause and the effect. Inferential accounts, by contrast, suggest that causal judgments are reasoning based rather than associative in nature. We used a direct measure of associative strength, identification of the outcome with which a cause was paired (cued recall), to see whether associative strength translated directly into causal ratings. Causal compounds AB+ and CD+ were intermixed with A+ and C- training. Cued-recall performance was better for cue B than for cue D; thus, associative strength was inherited by cue B from the strongly associated cue A (augmentation). However, the reverse was observed on the causal judgment measure: Cue B was judged to be less causal than D (cue competition). These results support an inferential over an associative account of causal judgments.     

Aging and retrospective revaluation of causal learning

In a 2-stage causal learning task, young and older participants first learned which foods presented in compound were followed by an allergic reaction (e.g., STEAK-BEANS→ REACTION) and then the causal efficacy of 1 food from these compounds was revalued (e.g., BEANS→ NO REACTION). In Experiment 1, unrelated food pairs were used, and although there were no age differences in compound- or single-cue-outcome learning, older adults did not retrospectively revalue the causal efficacy of the absent target cues (e.g., STEAK). However, they had weaker within-compound associations for the unrelated foods, and this may have prevented them from retrieving the representations of these cues. In Experiment 2, older adults still showed no retrospective revaluation of absent cues even though compound food cues with pre-existing associations were used (e.g., STEAK-POTATO), and they received additional learning trials. Finally, in Experiment 3, older adults revalued the causal efficacy of the target cues when small, unobtrusive icons of these cues were present during single-cue revaluation. These findings suggest that age-related deficits in causal learning for absent cues are due to ineffective associative binding and reactivation processes.     

Blocking and unblocking in human causal learning

Three experiments sought to develop the suggestion that, under some circumstances, common associative learning mechanisms might underlie animal conditioning and human causal learning, by demonstrating, in humans, an effect analogous to the unblocking by reinforcer omission observed in animal conditioning. Experiment 1 found no such effect. Experiment 2, designed to prevent inhibitory influences that might have masked excitatory unblocking in Experiment 1, demonstrated unblocking, indicating common human-animal associative learning mechanisms in which the associability of a stimulus varies as a function of its predictive history. Experiment 3, using a similar design but with a procedure promoting application of rational inference processes, failed to detect the same unblocking effect, indicating that associative and cognitive mechanisms may influence human causal learning.     

The tight coupling between category and causal learning

The main goal of the present research was to demonstrate the interaction between category and causal induction in causal model learning. We used a two-phase learning procedure in which learners were presented with learning input referring to two interconnected causal relations forming a causal chain (Experiment 1) or a common-cause model (Experiments 2a, b). One of the three events (i.e., the intermediate event of the chain, or the common cause) was presented as a set of uncategorized exemplars. Although participants were not provided with any feedback about category labels, they tended to induce categories in the first phase that maximized the predictability of their causes or effects. In the second causal learning phase, participants had the choice between transferring the newly learned categories from the first phase at the cost of suboptimal predictions, or they could induce a new set of optimally predictive categories for the second causal relation, but at the cost of proliferating different category schemes for the same set of events. It turned out that in all three experiments learners tended to transfer the categories entailed by the first causal relation to the second causal relation.     

Failures of explaining away and screening off in described versus experienced causal learning scenarios

Causal Bayes nets capture many aspects of causal thinking that set them apart from purely associative reasoning. However, some central properties of this normative theory routinely violated. In tasks requiring an understanding of explaining away and screening off, subjects often deviate from these principles and manifest the operation of an associative bias that we refer to as the rich-get-richer principle. This research focuses on these two failures comparing tasks in which causal scenarios are merely described (via verbal statements of the causal relations) versus experienced (via samples of data that manifest the intervariable correlations implied by the causal relations). Our key finding is that we obtained stronger deviations from normative predictions in the described conditions that highlight the instructed causal model compared to those that presented data. This counterintuitive finding indicate that a theory of causal reasoning and learning needs to integrate normative principles with biases people hold about causal relations.     

The outcome specificity of learned predictiveness effects: parallels between human causal learning and animal conditioning

Two experiments examined the outcome specificity of a learned predictiveness effect in human causal learning. Experiment 1 indicated that prior experience of a cue-outcome relation modulates learning about that cue with respect to a different outcome from the same affective class but not with respect to an outcome from a different affective class. Experiment 2 ruled out an interpretation of this effect in terms of context specificity. These results indicate that learned predictiveness effects in human causal learning index an associability that is specific to a particular class of outcomes. Moreover, they mirror demonstrations of the reinforcer specificity of analogous effects in animal conditioning, supporting the suggestion that, under some circumstances, human causal learning and animal conditioning reflect the operation of common associative mechanisms.     

Predictive versus diagnostic causal learning: Evidence from an overshadowing paradigm

Causal directionality belongs to one of the most fundamental aspects of causality that cannot be reduced to mere covariation. This paper is part of a debate between proponents of associative theories, which claim that learners are insensitive to the causal status of cues and outcomes, and proponents of causal-model theory, which postulates an interaction of assumptions about causal directionality and learning. Some researchers endorsing the associationist view have argued that evidence for the interaction between cue competition and causal directionality may be restricted to two-phase blocking designs. Furthermore, from the viewpoint of causal-model theory, blocking designs carry the potential problem that the predicted asymmetries of cue competition are partly dependent on asymmetries of retrospective inferences. The present experiments use a one-phase overshadowing paradigm that does not allow for retrospective inferences and therefore represents a more unambiguous test of sensitivity to causal directionality. The results strengthen causal-model theory by clearly demonstrating the influence of causal directionality on learning. However, they also provide evidence for boundary conditions for this effect by highlighting the role of the semantics of the learning task.     

Comparing positive and negative patterning in human learning

In human causal learning, positive patterning (PP) and negative patterning (NP) discriminations are often acquired at roughly the same rate, whereas PP is learned faster than NP in most experiments with nonhuman animals. One likely reason for this discrepancy is that most causal learning scenarios encourage participants to treat the presentation and omission of the relevant outcome as two events of comparable significance and likelihood. To investigate this, the current experiments compared PP and NP using a predictive learning paradigm based on a mock gambling task. In Experiment 1, one outcome (winning) was made more salient by being less frequent than the alternative outcome (losing). Under these circumstances, PP was learned faster than NP. In Experiment 2, subjects learned two PP and two NP discriminations, one involved win versus no change outcomes, the other involved lose versus no change outcomes. The subjects learned PP faster than NP, but only when discriminating win from no change. We argue that a difference in difficulty between PP and NP relies on a difference in the salience of the outcomes, consistent with the predictions of a relatively simple model of associative learning.     

Comparing positive and negative patterning in human learning

In human causal learning, positive patterning (PP) and negative patterning (NP) discriminations are often acquired at roughly the same rate, whereas PP is learned faster than NP in most experiments with nonhuman animals. One likely reason for this discrepancy is that most causal learning scenarios encourage participants to treat the presentation and omission of the relevant outcome as two events of comparable significance and likelihood. To investigate this, the current experiments compared PP and NP using a predictive learning paradigm based on a mock gambling task. In Experiment 1, one outcome (winning) was made more salient by being less frequent than the alternative outcome (losing). Under these circumstances, PP was learned faster than NP. In Experiment 2, subjects learned two PP and two NP discriminations, one involved win versus no change outcomes, the other involved lose versus no change outcomes. The subjects learned PP faster than NP, but only when discriminating win from no change. We argue that a difference in difficulty between PP and NP relies on a difference in the salience of the outcomes, consistent with the predictions of a relatively simple model of associative learning.     

A dissociation between causal judgement and the ease with which a cause is categorized with its effect

The associative view of human causal learning argues that causation is attributed to the extent that the putative cause activates, via an association, a mental representation of the effect. That is, causal learning is a human analogue of animal conditioning. We tested this associative theory using a task in which a fictitious character suffered from two allergic reactions, rash (O₁) and headache (O₂). In a conditioned inhibition design with each of these two outcomes (A-O₁/AX- and B-O₂/BY-), participants were trained that one herbal remedy (X) prevented O₁ and that the other (Y) prevented O₂. These inhibitory properties were revealed in a causal judgement summation test. In a subsequent categorization task, X was most easily categorized with O_1, and Y with O₂. Thus, the categorization data indicated an excitatory X-O₁ and Y-O₂ association, the reverse of the inhibitory relationship observed on the causal judgement measure. A second experiment showed that this pattern of excitation and inhibition is dependent on intermixed A-O₁ and AX- trials. These results are problematic for the standard application of associative activation theories to causal judgement. We argue instead that the inhibition revealed in the causal judgement task reflects inferential reasoning, which relies, in part, on the ability of the cue in question to excite a representation of the outcome, as revealed in the categorization test.     

Re-assessing causal accounts of learnt behavior in rats

Rats received either a common-cause (i.e., A→B, A→food) or a causal-chain training scenario (i.e., B→A, A→food) before their tendency to approach the food magazine during the presentation of B was assessed as a function of whether it was preceded by a potential alternative cause. Causal model theory predicts that the influence of an alternative cause should be restricted to the common-cause scenario. In Experiment 1, responding to B was reduced when it occurred after pressing a novel lever during the test phase. This effect was not influenced by the type of training scenario. In Experiment 2, rats were familiarized with the lever prior to test by training it as a potential cause of B. After this treatment, the lever now failed to influence test responding to B. In Experiment 3, rats given common-cause training responded more to B when it followed a cue that had previously been trained as a predictor of B, than when it followed another stimulus. This effect was not apparent in rats that received causal-chain training. This pattern of results is the opposite of that predicted by causal model theory. Thus, in three experiments, the presence of an alternative cause failed to influence test responding in manner consistent with causal model theory. These results undermine the application of causal model theory to rats, but are consistent with associative analyses.     

Independence and dependence in human causal reasoning

Causal graphical models (CGMs) are a popular formalism used to model human causal reasoning and learning. The key property of CGMs is the causal Markov condition, which stipulates patterns of independence and dependence among causally related variables. Five experiments found that while adult’s causal inferences exhibited aspects of veridical causal reasoning, they also exhibited a small but tenacious tendency to violate the Markov condition. They also failed to exhibit robust discounting in which the presence of one cause as an explanation of an effect makes the presence of another less likely. Instead, subjects often reasoned “associatively,” that is, assumed that the presence of one variable implied the presence of other, causally related variables, even those that were (according to the Markov condition) conditionally independent. This tendency was unaffected by manipulations (e.g., response deadlines) known to influence fast and intuitive reasoning processes, suggesting that an associative response to a causal reasoning question is sometimes the product of careful and deliberate thinking. That about 60% of the erroneous associative inferences were made by about a quarter of the subjects suggests the presence of substantial individual differences in this tendency. There was also evidence that inferences were influenced by subjects’ assumptions about factors that disable causal relations and their use of a conjunctive reasoning strategy. Theories that strive to provide high fidelity accounts of human causal reasoning will need to relax the independence constraints imposed by CGMs.     

Creators' intentions bias judgments of function independently from causal inferences

Participants learned about novel artifacts that were created for function X, but later used for function Y. When asked to rate the extent to which X and Y were a given artifact's function, participants consistently rated X higher than Y. In Experiments 1 and 2, participants were also asked to rate artifacts' efficiency to perform X and Y. This allowed us to test if participants' preference for X was mediated by causal inferences. Experiment 1 showed that participants did not infer intentionally created artifacts performed X more efficiently than Y. Experiment 2 showed participants did not infer that only an efficient (but not an inefficient) artifact provided evidence of intentional creation. Causal inferences involving efficiency, did not account for participants' preferences. In Experiment 3, in contrast, when the creator changed her mind about an artifact's function (i.e., from X to Y), the preference for the original function tended to disappear. Creators' intentions were the basis for participants' preference. Results are discussed relative to essentialist theories.     

Structure and strength in causal induction

We present a framework for the rational analysis of elemental causal induction-learning about the existence of a relationship between a single cause and effect-based upon causal graphical models. This framework makes precise the distinction between causal structure and causal strength: the difference between asking whether a causal relationship exists and asking how strong that causal relationship might be. We show that two leading rational models of elemental causal induction, DeltaP and causal power, both estimate causal strength, and we introduce a new rational model, causal support, that assesses causal structure. Causal support predicts several key phenomena of causal induction that cannot be accounted for by other rational models, which we explore through a series of experiments. These phenomena include the complex interaction between DeltaP and the base-rate probability of the effect in the absence of the cause, sample size effects, inferences from incomplete contingency tables, and causal learning from rates. Causal support also provides a better account of a number of existing datasets than either DeltaP or causal power.     

When learning order affects sensitivity to base rates: challenges for theories of causal learning

In three experiments we investigated whether two procedures of acquiring knowledge about the same causal structure, predictive learning (from causes to effects) versus diagnostic learning (from effects to causes), would lead to different base-rate use in diagnostic judgments. Results showed that learners are capable of incorporating base-rate information in their judgments regardless of the direction in which the causal structure is learned. However, this only holds true for relatively simple scenarios. When complexity was increased, base rates were only used after diagnostic learning, but were largely neglected after predictive learning. It could be shown that this asymmetry is not due to a failure of encoding base rates in predictive learning because participants in all conditions were fairly good at reporting them. The findings present challenges for all theories of causal learning.     

Accessing causal relations in semantic memory

Most studies investigating semantic memory have focused on taxonomic or associative relations. Little is known about how other relations, such as causal relations, are represented and accessed. In three experiments, we presented participants with pairs of words one after another, describing events that referred to either a cause (e.g., spark) or an effect (e.g., fire). We manipulated the temporal order of word presentation and the question participants had to respond to. The results revealed that questions referring to the existence of a causal relation are answered faster when the first word refers to a cause and the second word refers to its effect than vice versa. However, no such asymmetry was observed with questions referring to the associative relation. People appear to distinguish the roles of cause and effect when queried specifically about a causal relation, but not when the same information is evaluated for the presence of an associative relation.     

Seeing versus doing: two modes of accessing causal knowledge

The ability to derive predictions for the outcomes of potential actions from observational data is one of the hallmarks of true causal reasoning. We present four learning experiments with deterministic and probabilistic data showing that people indeed make different predictions from causal models, whose parameters were learned in a purely observational learning phase, depending on whether learners believe that an event within the model has been merely observed ("seeing") or was actively manipulated ("doing"). The predictions reflect sensitivity both to the structure of the causal models and to the size of their parameters. This competency is remarkable because the predictions for potential interventions were very different from the patterns that had actually been observed. Whereas associative and probabilistic theories fail, recent developments of causal Bayes net theories provide tools for modeling this competency.     

PI and RI in serial learning as a function of environmental context

The influence of the environmental context upon serial learning was investigated in a PI design in Experiment 1 and an RI design in Experiment 2. Either one or four lists learned either before or after the critical list were used to manipulate PI or RI, respectively. Learning the critical list in the same room as the interference-inducing lists or in a different room provided the first context manipulation. The second context factor involved relearning the critical list in the same room as it was learned 24 hours earlier, or in a different one. In the PI study the early and middle thirds of the list were affected by context in original learning. In relearning, correct responses over the first two trials differed both as a function of number of prior lists learned and of original learning context. With RI, fewer trials to relearn were required by a condition involving facilitating context manipulations, relative to a neutral context, and a competing context condition was inferior. The results are largely consistent with predictions derived from the interference theory of forgetting and traditional associative learning theory.     

Inference-based retrospective revaluation in human causal judgments requires knowledge of within-compound relationships

In an allergist causal-judgment task, food compounds were followed by an allergic reaction (e.g., AB+), and then 1 cue (A) was revalued. Experiment 1, in which participants who were instructed that whatever was true about one element of a causal compound was also true of the other, showed a reverse of the standard retrospective revaluation effect. That is, ratings of B were higher when A was causal (A+) than when A was safe (A-). This effect was taken to reflect inferential reasoning, not an associative mechanism. In Experiment 2, within-compound associations were found to be necessary to produce this inference-based revaluation. Therefore, evidence that within-compound associations are necessary for retrospective revaluation is consistent with the inferential account of causal judgments.