Computation and Causation

The computer's effect on our understanding of causation has been enormous. By the mid-1980s, philosophical and social-scientific work on the topic had left us with (1) no reasonable reductive account of causation and (2) a class of statistical causal models tied to linear regression. At this time, computer scientists were attacking the problem of equipping robots with models of the external that included probabilistic portrayals of uncertainty. To solve the problem of efficiently storing such knowledge, they introduced Bayes Networks and directed graphs. By attaching a causal interpretation to Bayes Networks, the philosophy of causation changed dramatically. We are now able to be extremely general about how causal structure connects to data, and systematic about when causal structures are empirically indistinguishable. In this essay I try to motivate and describe this synthesis.     

The Impossibility of Emergent Conscious Causal Powers

This paper argues that emergent conscious properties can't bestow emergent causal powers. It supports this conclusion by way of a dilemma. Necessarily, an (allegedly efficacious) emergent conscious property brings about its effects actively or other than actively (in senses explained in the paper). If actively, then, the paper argues, the emergent conscious property can't have causal powers at all. And if other than actively, then, the paper argues, the emergentist finds himself committed to incompatible accounts of causation.     

Beyond the Information Given: Causal Models in Learning and Reasoning

ABSTRACT— The philosopher David Hume's conclusion that causal induction is solely based on observed associations still presents a puzzle to psychology. If we only acquired knowledge about statistical covariations between observed events without accessing deeper information about causality, we would be unable to understand the differences between causal and spurious relations, between prediction and diagnosis, and between observational and interventional inferences. All these distinctions require a deep understanding of causality that goes beyond the information given. We report a number of recent studies that demonstrate that people and rats do not stick to the superficial level of event covariations but reason and learn on the basis of deeper causal representations. Causal-model theory provides a unified account of this remarkable competence.     

The second offspring of General Process Learning Theory: Overt behavior as the ambassador of the mind

Wasserman suggested in a recent book review that the study of intervening cognitive processes represents a current focus of interest in animal learning and that this has led to a revitalization of comparative psychology. An examination of the volume reviewed suggests that he may have overstated the case. Most of the authors to whom he refers expressed dissatisfaction with traditional stimulus-response associationism but few argued for the extreme (information processing) sort of cognitive approach described by Wasserman.     

The Accuracy of Causal Learning Over Long Timeframes: An Ecological Momentary Experiment Approach

The ability to learn cause–effect relations from experience is critical for humans to behave adaptively — to choose causes that bring about desired effects. However, traditional experiments on experience-based learning involve events that are artificially compressed in time so that all learning occurs over the course of minutes. These paradigms therefore exclusively rely upon working memory. In contrast, in real-world situations we need to be able to learn cause–effect relations over days and weeks, which necessitates long-term memory. 413 participants completed a smartphone study, which compared learning a cause–effect relation one trial per day for 24 days versus the traditional paradigm of 24 trials back- to- back. Surprisingly, we found few differences between the short versus long timeframes. Subjects were able to accurately detect generative and preventive causal relations, and they exhibited illusory correlations in both the short and long timeframe tasks. These results provide initial evidence that experience-based learning over long timeframes exhibits similar strengths and weaknesses as in short timeframes. However, learning over long timeframes may become more impaired with more complex tasks.     

Causal Necessity in Aristotle

Like many realists about causation and causal powers, Aristotle uses the language of necessity when discussing causation, and he appears to think that by invoking necessity, he is clarifying the manner in which causes bring about or determine their effects. In so doing, he would appear to run afoul of Humean criticisms of the notion of a necessary connection between cause and effect. The claim that causes necessitate their effects may be understood— or attacked— in several ways, however, and so whether the view or its criticism is tenable depends on how we understand the necessitation claim. In fact, Aristotelian efficient causation may be said to involve two distinct necessary connections: one is a relation between causes considered as potential, while the other relates them considered as active. That is, the claims that (1) what has the power to heat necessarily heats what has the power to be heated, and that (2) a particular flame which is actually under a pot necessarily heats it, both of which appear to be true for Aristotle, involve distinct notions of necessity. The latter kind of necessity is based on the facts, as Aristotle sees them, about change, whereas the former is based in the nature of properties. Though different, both kinds of necessity are instances of what contemporary philosophers would call metaphysical necessity, and together they also amount to a theory of causal determination.     

Thorndike's Legacy: Learning, Selection, And The Law Of Effect

This introduction to a symposium on the centennial of Edward L. Thorndike's 1898 monograph on animal intelligence briefly considers the origins of his law of effect and the influence of Darwin's selectionism. It also provides the background for an unfinished book review by William W. Cumming of a biography of Thorndike. The review places in historical context Thorndike's position both on psychology as a science of behavior and on the vocabulary of that science.     

Augmentation: Synergistic Conditioning in Taste-Aversion Learning

Recent work in taste-aversion learning has revealed a new phenomenon in classical conditioning. When a preconditioned gustatory cue (taste or odor) is conditioned in compound with a second gustatory cue, conditioning to the second cue is augmented. This enhanced conditioning of the second cue is noteworthy because studies with other forms of classical conditioning have shown blocked conditioning to the second cue. This new phenomenon has been termed augmentation , and it has implications for the study of taste and odor interactions, formal models of learning, and clinical interventions with cancer patients.     

Learning to fear a second-order stimulus following vicarious learning

Vicarious fear learning refers to the acquisition of fear via observation of the fearful responses of others. The present study aims to extend current knowledge by exploring whether second-order vicarious fear learning can be demonstrated in children. That is, whether vicariously learnt fear responses for one stimulus can be elicited in a second stimulus associated with that initial stimulus. Results demonstrated that children's (5–11 years) fear responses for marsupials and caterpillars increased when they were seen with fearful faces compared to no faces. Additionally, the results indicated a second-order effect in which fear-related learning occurred for other animals seen together with the fear-paired animal, even though the animals were never observed with fearful faces themselves. Overall, the findings indicate that for children in this age group vicariously learnt fear-related responses for one stimulus can subsequently be observed for a second stimulus without it being experienced in a fear-related vicarious learning event. These findings may help to explain why some individuals do not recall involvement of a traumatic learning episode in the development of their fear of a specific stimulus.     

Communal Versus Individual Learning of a Math-Estimation Task: African American Children and the Culture of Learning Contexts

The authors compared the performance of 78 African American 5th-grade students who studied a math-estimation task in one of two learning contexts. Learning contexts differed in the degree to which they afforded the expression of communalism. ANCOVA confirmed that posttest performance was best for students who studied in the high communal-learning context. The findings support A. W. Boykin's (1994) contention that the cultural context of learning can be a critical mediator of children's performance.     

Timing at the Start of Associative Learning

Some theories of associative learning imply that time plays a fundamental role in the acquisition process. Consistent with these theories, this paper presents evidence that the time from the onset of a conditioned stimulus (CS) until presentation of the unconditioned stimulus (US) is learned very rapidly at the start of training. We report two autoshaping studies and a study on aversive conditioning in goldfish in which we examine timing at the start of conditioning. We also review data from a number of other conditioning preparations, including fear-potentiated startle, appetitive conditioning in rats, and eyeblink conditioning in rabbits, that report conditioned response (CR) timing early in training. Acquisition speed and the very first expressions of conditioned responding often show sensitivity to the time of US presentation. In instances where temporal control is slowly expressed, it is likely due to performance factors, not to slow learning about time. In fact, the learning about time may be a necessary condition for associative learning.     

Operant Reward Learning in Aplysia

Anticipating the future has a decided evolutionary advantage, and researchers have found many evolutionarily conserved mechanisms by which humans and animals learn to predict future events. Researchers often study such learned behavior using conditioning experiments. The marine snail Aplysia has been at the forefront of research into the cellular and molecular mechanisms of classical conditioning. Recently, Aplysia has also gained a reputation as a valuable model system for operant reward learning. Its feeding behavior can be operantly conditioned in the intact animal as well as in reduced preparations of the nervous system. The reward signal relies on dopamine transmission and acts in conjunction with activity in an identified neuron (B51) to bring about operant memory.     

A Bayesian Theory of Sequential Causal Learning and Abstract Transfer

Two key research issues in the field of causal learning are how people acquire causal knowledge when observing data that are presented sequentially, and the level of abstraction at which learning takes place. Does sequential causal learning solely involve the acquisition of specific cause‐effect links, or do learners also acquire knowledge about abstract causal constraints? Recent empirical studies have revealed that experience with one set of causal cues can dramatically alter subsequent learning and performance with entirely different cues, suggesting that learning involves abstract transfer, and such transfer effects involve sequential presentation of distinct sets of causal cues. It has been demonstrated that pre‐training (or even post‐training) can modulate classic causal learning phenomena such as forward and backward blocking. To account for these effects, we propose a Bayesian theory of sequential causal learning. The theory assumes that humans are able to consider and use several alternative causal generative models, each instantiating a different causal integration rule. Model selection is used to decide which integration rule to use in a given learning environment in order to infer causal knowledge from sequential data. Detailed computer simulations demonstrate that humans rely on the abstract characteristics of outcome variables (e.g., binary vs. continuous) to select a causal integration rule, which in turn alters causal learning in a variety of blocking and overshadowing paradigms. When the nature of the outcome variable is ambiguous, humans select the model that yields the best fit with the recent environment, and then apply it to subsequent learning tasks. Based on sequential patterns of cue‐outcome co‐occurrence, the theory can account for a range of phenomena in sequential causal learning, including various blocking effects, primacy effects in some experimental conditions, and apparently abstract transfer of causal knowledge.     

The Causal Structure of Evolutionary Theory

One contentious debate in the philosophy of biology is that between the statisticalists and causalists. The former understand core evolutionary concepts like fitness and selection to be mere statistical summaries of underlying causal processes. In this view, evolutionary changes cannot be causally explained by selection or fitness. The causalist side, on the other hand, holds that populations can change in response to selection—one can cite fitness differences or driftability in causal explanations of evolutionary change. But, on the causalist side, it is often not clear how, precisely, one should understand these causes. Thus, much more could be said about what sort of causes fitness and driftability are. In this paper, I borrow Dretske's distinction between structuring and triggering causes and I suggest that fitness and driftability are structuring causes of evolution.     

Causal theories of the moving spotlight

This paper brings together the Sarvāstivāda (a major school of Abhidharma Buddhism) and Miller's (2019) moving spotlight theory to see how presentness is explained in terms of causation. The paper argues that a causal theory of presentness like Miller's encounters a dilemma: causation is either synchronic or diachronic, but neither is safe in the presence of the challenges. On the one hand, if causation is synchronic, how does a causal chain extend over time so that the wave of causation (and hence the spotlight) can move at all? On the other, if causation is diachronic, then a present moment has a temporal extension enough to accommodate a cause and its proximate effect. The paper proposes a synchronic notion of causation that counters the first challenge. This can be done by bringing together the Sarvāstivādin grasp-production model of causation, some Stoic ideas about causation, and some Armstrong's metaphysics.     

There are Non‐Causal Explanations of Particular Events

A defence of non‐causal explanations of events is presented in cases where explanation is understood as modal explanation. In such cases the source of modal information is crucial. All explanations implicitly use contrast classes, and relative to a particular contrast we can privilege some difference makers over others. Thinking about changes in these privileged “actual” difference makers is then the source of modal information for any given explanandum. If an actual difference maker is non‐causal, then we have a principled definition of a non‐causal explanation of an event regardless of how much causal information is also contained in the explanans. A demarcation of explanation into causal and non‐causal in this way recovers ordinary language about explanation as well as reflecting genuine differences in practice, such as the in‐principle evidential base for any modal claim.     

Causal Inference from Descriptions of Experimental and Non-Experimental Research: Public Understanding of Correlation-Versus-Causation

The human tendency to conflate correlation with causation has been lamented by various scientists (Kida, 2006; Stanovich, 2009), and vivid examples of it can be found in both the media and peer-reviewed literature. However, there is little systematic data on the extent to which individuals conflate correlation with causation. In three experiments, we presented people with one of four research vignettes generated from the combination of two independent variables: whether the vignette described an experimental or non-experimental design, and whether it revealed a positive or negative association. Upon reading their vignette, participants selected inferences that could be drawn from the findings. Participants drew causal inferences from non-experimental vignettes as often as they did from experimental vignettes, and more frequently for causal statements and directions of association that fit with intuitive notions than for those that did not. We discuss our findings in relation to other biases in human thinking.     

The Learning Signal in Perceptual Tuning of Speech: Bottom Up Versus Top-Down Information

Cognitive systems face a tension between stability and plasticity. The maintenance of long-term representations that reflect the global regularities of the environment is often at odds with pressure to flexibly adjust to short-term input regularities that may deviate from the norm. This tension is abundantly clear in speech communication when talkers with accents or dialects produce input that deviates from a listener's language community norms. Prior research demonstrates that when bottom-up acoustic information or top-down word knowledge is available to disambiguate speech input, there is short-term adaptive plasticity such that subsequent speech perception is shifted even in the absence of the disambiguating information. Although such effects are well-documented, it is not yet known whether bottom-up and top-down resolution of ambiguity may operate through common processes, or how these information sources may interact in guiding the adaptive plasticity of speech perception. The present study investigates the joint contributions of bottom-up information from the acoustic signal and top-down information from lexical knowledge in the adaptive plasticity of speech categorization according to short-term input regularities. The results implicate speech category activation, whether from top-down or bottom-up sources, in driving rapid adjustment of listeners' reliance on acoustic dimensions in speech categorization. Broadly, this pattern of perception is consistent with dynamic mapping of input to category representations that is flexibly tuned according to interactive processing accommodating both lexical knowledge and idiosyncrasies of the acoustic input.