Negative causation in causal and mechanistic explanation

Instances of negative causation—preventions, omissions, and the like—have long created philosophical worries. In this paper, I argue that concerns about negative causation can be addressed in the context of causal explanation generally, and mechanistic explanation specifically. The gravest concern about negative causation is that it exacerbates the problem of causal promiscuity—that is, the problem that arises when a particular account of causation identifies too many causes for a particular effect. In the explanatory context, the problem of promiscuity can be solved by characterizing the phenomenon to be explained as a contrast between two or more events or non-events. This contrastive strategy also can solve other problems that negative causation presents for the leading accounts of mechanistic explanation. Along the way, I argue that to be effective, accounts of causal explanation must incorporate negative causation. I also develop a taxonomy of negative causation and incorporate each variety of negative causation into the leading accounts of mechanistic explanation.     

From covariation to causation: a test of the assumption of causal power

How humans infer causation from covariation has been the subject of a vigorous debate, most recently between the computational causal power account (P. W. Cheng, 1997) and associative learning theorists (e.g., K. Lober & D. R. Shanks, 2000). Whereas most researchers in the subject area agree that causal power as computed by the power PC theory offers a normative account of the inductive process. Lober and Shanks, among others, have questioned the empirical validity of the theory. This article offers a full report and additional analyses of the original study featured in Lober and Shanks's critique (M. J. Buehner & P. W. Cheng, 1997) and reports tests of Lober and Shanks's and other explanations of the pattern of causal judgments. Deviations from normativity, including the outcome-density bias, were found to be misperceptions of the input or other artifacts of the experimental procedures rather than inherent to the process of causal induction.     

Temporal delays can facilitate causal attribution: Towards a general timeframe bias in causal induction

Does reasoning occur on the Wason selection task, or are card selections determined purely on the basis of heuristic processes? To answer this question two relevance-based theories of reasoning are compared: (1) the theory of Evans (1984, 1989; Evans &; Over, 1996), which takes the heuristic viewpoint, and (2) the theory of Sperber, Cara, and Girotto (1995), which takes the reasoning viewpoint. In three experiments, the effect of removing matching cards from the selection task array is examined. It is argued that the Sperber et al. theory makes clearer predictions about the results of these manipulations, which are confirmed, and that the Evans theory can only accommodate them if it allows the operation of reasoning processes. The results are also discussed in relation to Roth's (1979) account of the selection task, mental models theory, and information gain theory.     

Covariation in natural causal induction.

The covariation component of everyday causal inference has been depicted, in both cognitive and social psychology as well as in philosophy, as heterogeneous and prone to biases. The models and biases discussed in these domains are analyzed with respect to focal sets: contextually determined sets of events over which covariation is computed. Moreover, these models are compared to our probabilistic contrast model, which specifies causes as first and higher order contrasts computed over events in a focal set. Contrary to the previous depiction of covariation computation, the present assessment indicates that a single normative mechanism--the computation of probabilistic contrasts--underlies this essential component of natural causal induction both in everyday and in scientific situations.     

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.     

Encouraging children to think counterfactually enhances blocking in a causal learning task

According to a higher order reasoning account, inferential reasoning processes underpin the widely observed cue competition effect of blocking in causal learning. The inference required for blocking has been described as modus tollens (if p then q, not q therefore not p). Young children are known to have difficulties with this type of inference, but research with adults suggests that this inference is easier if participants think counterfactually. In this study, 100 children (51 five-year-olds and 49 six- to seven-year-olds) were assigned to two types of pretraining groups. The counterfactual group observed demonstrations of cues paired with outcomes and answered questions about what the outcome would have been if the causal status of cues had been different, whereas the factual group answered factual questions about the same demonstrations. Children then completed a causal learning task. Counterfactual pretraining enhanced levels of blocking as well as modus tollens reasoning but only for the younger children. These findings provide new evidence for an important role for inferential reasoning in causal learning.     

Risk strategy under task complexity: a multivariate analysis of behaviour in a naturalistic setting

This study complements the existing literature on decision-making processes and outcomes in complex settings by exploring the impact of different types of complexity on risk strategies in a naturalistic setting. The study analyses a large sample of decisions made by individuals in UK offcourse betting markets, a fertile environment for observing both a variety of risk strategies and a range of task complexities. Specifically, the investigation focuses on the comparative impacts of complexity defined in terms, respectively, of alternatives and attributes. The results suggest that the risk strategy employed is affected by task complexity. Complexity does not affect the size of risk accepted but alternative- and attribute-based complexity together influence the propensity to accept greater degrees of risk. In addition, the effect of attribute-based complexity on risk taking appears to be modified by the use of risk-hedging strategies. The results are observed to corroborate some earlier work on decision process and outcome; where differences with earlier findings are identified, some possible explanations are offered. © 1998 John Wiley & Sons, Ltd.     

Associative and causal reasoning accounts of causal induction: Symmetries and asymmetries in predictive and diagnostic inferences

Associative and causal reasoning accounts are probably the two most influential types of accounts of causal reasoning processes. Only causal reasoning accounts predict certain asymmetries between predictive (i.e., reasoning from causes to effects) and diagnostic (i.e., reasoning from effects to causes) inferences regarding cue-interaction phenomena (e.g., the overshadowing effect). In the experiments reported here, we attempted to delimit the conditions under which these asymmetries occur. The results show that unless participants perceived the relevance of causal information to solving the task, predictive and diagnostic inferences were symmetrical. Specifically, Experiments 1A and 1B showed that implicitly stressing the relevance of causal information by having participants review the instructions favored the presence of asymmetries between predictive and diagnostic situations. In addition, Experiment 2 showed that explicitly stressing the relevance of causal information by stating the importance of the causal role of events after the instructions were given also favored the asymmetry.     

The experimental task influences cue competition in human causal learning

Participants were shown A+ and C- trials followed by AB+ and CD+ trials. These trials were embedded in a causal learning task in which participants had to learn either the relationship between different foods and allergic reactions or the relationship between different stocks and an increase in the stock market index. The authors orthogonally varied the manner in which the different cues were presented to participants during training. Cue competition was related to the causal learning scenario but not to the manner in which the different cues were presented. These results question claims of a human bias toward configural processing that were based on difficulties in finding cue competition in some previous causal learning experiments.     

Human causal induction: A glimpse at the whole picture

In the present work, most relevant evidence in causal learning literature is reviewed and a general cognitive architecture based on the available corpus of experimental data is proposed. However, contrary to algorithms formulated in the Bayesian nets framework, such architecture is not assumed to optimise the usefulness of the available information in order to induce the underlying causal structure as a whole. Instead, human reasoners seem to rely heavily on local clues and previous knowledge to discriminate between spurious and truly causal covariations, and piece those relations together only when they are demanded to do so. Bayesian networks and AI algorithms for causal inference are nonetheless considered valuable tools to identify the main computational goals of causal induction processes and to define the problems any intelligent causal inference system must solve.     

Magical thinking in judgments of causation: Can anomalous phenomena affect ontological causal beliefs in children and adults?

In four experiments, 4‐, 5‐, 6‐ and 9‐year‐old children and adults were tested on the entrenchment of their magical beliefs and their beliefs in the universal power of physical causality. In Experiment 1, even 4‐year‐olds showed some understanding of the difference between ordinary and anomalous (magical) causal events, but only 6‐year‐olds and older participants denied that magic could occur in real life. When shown an anomalous causal event (a transformation of a physical object in an apparently empty box after a magic spell was cast on the box), 4‐ and 6‐year‐olds accepted magical explanations of the event, whereas 9‐year‐olds and adults did not. In Experiment 2, the same patterns of behaviour as above were shown by 6‐ and 9‐year‐olds who demonstrated an understanding of the difference between genuine magical events and similarly looking tricks. Testing the entrenchment of magical beliefs in this experiment showed that 5‐year‐olds tended to retain their magical explanations of the anomalous event, even after the mechanism of the trick had been explained to them, whereas 6‐ and 9‐year‐olds did not. In Experiment 3, adult participants refused to accept magical explanations of the anomalous event and interpreted it as a trick or an illusion, even after this event was repeated 4 times. Yet, when in Experiment 4 similar anomalous causal events were demonstrated without reference to magic, most adults acknowledged, both in their verbal judgments and in their actions, that the anomalous effects were not a fiction but had really occurred. The data of this study suggest that in the modern industrialized world, magical beliefs persist but are disguised to fit the dominant scientific paradigm.     

Experimental setting affects the performance of guppies in a numerical discrimination task

A recent study found that guppies (Poecilia reticulata) can be trained to discriminate 4 versus 5 objects, a numerical discrimination typically achieved only by some mammals and birds. In that study, guppies were required to discriminate between two patches of small objects on the bottom of the tank that they could remove to find a food reward. It is not clear whether this species possesses exceptional numerical accuracy compared with the other ectothermic vertebrates or whether its remarkable performance was due to a specific predisposition to discriminate between differences in the quality of patches while foraging. To disentangle these possibilities, we trained guppies to the same numerical discriminations with a more conventional two-choice discrimination task. Stimuli were sets of dots presented on a computer screen, and the subjects received a food reward upon approaching the set with the larger numerosity. Though the cognitive problem was identical in the two experiments, the change in the experimental setting led to a much poorer performance as most fish failed even the 2 versus 3 discrimination. In four additional experiments, we varied the duration of the decision time, the type of stimuli, the length of training, and whether correction was allowed in order to identify the factors responsible for the difference. None of these parameters succeeded in increasing the performance to the level of the previous study, although the group trained with three-dimensional stimuli learned the easiest numerical task. We suggest that the different results with the two experimental settings might be due to constraints on learning and that guppies might be prepared to accurately estimate patch quality during foraging but not to learn an abstract stimulus–reward association.     

Performance of a visuomotor walking task in an augmented reality training setting

Visual cues can be used to train walking patterns. Here, we studied the performance and learning capacities of healthy subjects executing a high-precision visuomotor walking task, in an augmented reality training set-up. A beamer was used to project visual stepping targets on the walking surface of an instrumented treadmill. Two speeds were used to manipulate task difficulty. All participants (n = 20) had to change their step length to hit visual stepping targets with a specific part of their foot, while walking on a treadmill over seven consecutive training blocks, each block composed of 100 stepping targets. Distance between stepping targets was varied between short, medium and long steps. Training blocks could either be composed of random stepping targets (no fixed sequence was present in the distance between the stepping targets) or sequenced stepping targets (repeating fixed sequence was present). Random training blocks were used to measure non-specific learning and sequenced training blocks were used to measure sequence-specific learning. Primary outcome measures were performance (% of correct hits), and learning effects (increase in performance over the training blocks: both sequence-specific and non-specific). Secondary outcome measures were the performance and stepping-error in relation to the step length (distance between stepping target). Subjects were able to score 76% and 54% at first try for lower speed (2.3 km/h) and higher speed (3.3 km/h) trials, respectively. Performance scores did not increase over the course of the trials, nor did the subjects show the ability to learn a sequenced walking task. Subjects were better able to hit targets while increasing their step length, compared to shortening it. In conclusion, augmented reality training by use of the current set-up was intuitive for the user. Suboptimal feedback presentation might have limited the learning effects of the subjects.     

Domestic cats (Felis catus) do not show causal understanding in a string-pulling task

This study explored how domestic cats perform in a horizontal string-pulling task to determine whether they understand this case of physical causality. Fifteen cats were tested on their ability to retrieve an unreachable food treat in three different set-ups: (a) a single baited string, (b) two parallel strings where only one was baited and (c) two crossed strings where only one was baited. All cats succeeded at pulling a single string to obtain a treat, but none consistently chose the correct string when two strings were parallel. When tested with two crossed strings one cat chose the wrong string consistently and all others performed at chance level. There was no evidence that cats understand the function of the strings or their physical causality.     

Dual frames for causal induction: the normative and the heuristic

Causal induction in the real world often has to be quick and efficient as well as accurate. We propose that people use two different frames to achieve these goals. The A-frame consists of heuristic processes that presuppose rarity and can detect causally relevant factors quickly. The B-frame consists of analytic processes that can be highly accurate in detecting actual causes. Our dual frame theory implies that several factors affect whether people use the A-frame or the B-frame in causal induction: among these are symmetrical negation, intervention and commitment. This theory is tested and sustained in two experiments. The results also provide broad support for dual process accounts of human thinking in general.