Marr's levels of analysis—computational, algorithmic, and implementation—have served cognitive science well over the last 30 years. But the recent increase in the popularity of the computational level raises a new challenge: How do we begin to relate models at different levels of analysis? We propose that it is possible to define levels of analysis that lie between the computational and the algorithmic, providing a way to build a bridge between computational- and algorithmic-level models. The key idea is to push the (...) notion of rationality, often used in defining computational-level models, deeper toward the algorithmic level. We offer a simple recipe for reverse-engineering the mind's cognitive strategies by deriving optimal algorithms for a series of increasingly more realistic abstract computational architectures, which we call “resource-rational analysis.”. (shrink)

Real causal systems are complicated. Despite this, causal learning research has traditionally emphasized how causal relations can be induced on the basis of idealized events, i.e. those that have been mapped to binary variables and abstracted from time. For example, participants may be asked to assess the efficacy of a headache-relief pill on the basis of multiple patients who take the pill (or not) and find their headache relieved (or not). In contrast, the current study examines learning via interactions with (...) continuous dynamic systems, systems that include continuous variables that interact over time (and that can be continuously observed in real time by the learner). To explore such systems, we develop a new framework that represents a causal system as a network of stationary Gauss--Markov (``\OU'') processes and show how such \textit{OU networks} can express complex dynamic phenomena such as feedback loops and oscillations. To assess adult's abilities to learn such systems, we conducted an experiment in which participants were asked to identify the causal relationships of a number of OU networks, potentially carrying out multiple, temporally-extended interventions. We compared their judgments to a normative model for learning OU networks as well as a range of alternative and heuristic learning models from the literature. We found that, although participants exhibited substantial learning of such systems, they committed certain systematic errors. These successes and failures were best accounted for by a model that describes people as focusing on pairs of variables, rather than evaluating the evidence with respect to the full space of possible structural models. We argue that our approach provides both a principled framework for exploring the space of dynamic learning environments as well as new algorithmic insights into how people interact successfully with a continuous causal world. (shrink)

How do we make causal judgments? Many studies have demonstrated that people are capable causal reasoners, achieving success on tasks from reasoning to categorization to interventions. However, less is known about the mental processes used to achieve such sophisticated judgments. We propose a new process model—the mutation sampler—that models causal judgments as based on a sample of possible states of the causal system generated using the Metropolis–Hastings sampling algorithm. Across a diverse array of tasks and conditions encompassing over 1,700 participants, (...) we found that our model provided a consistently closer fit to participant judgments than standard causal graphical models. In particular, we found that the biases introduced by mutation sampling accounted for people's consistent, predictable errors that the normative model by definition could not. Moreover, using a novel experimental methodology, we found that those biases appeared in the samples that participants explicitly judged to be representative of a causal system. We conclude by advocating sampling methods as plausible process‐level accounts of the computations specified by the causal graphical model framework and highlight opportunities for future research to identify not just what reasoners compute when drawing causal inferences, but also how they compute it. (shrink)

Two variables are usually recognised as determinants of human causal learning: the contingency between a candidate cause and effect, and the temporal and/or spatial contiguity between them. A common finding is that reductions in temporal contiguity produce concomitant decrements in causal judgement. This finding had previously (Shanks & Dickinson, 1987) been interpreted as evidence that causal induction is based on associative learning processes. Buehner and May (2002, 2003, 2004) have challenged this notion by demonstrating that the impact of temporal delay (...) depends on expectations about the timeframe between cause and effect. A corollary of this knowledge-mediation account is that in certain situations longer delays could facilitate, while shorter delays should impair, causal learning. Here we present two experiments involving a physical apparatus that demonstrate a detrimental effect of contiguity under certain conditions. In contrast to all previous studies, delays universally promoted causal learning. This evidence is clearly at variance with the notion of a bottom-up contiguity bias in causal induction. A new, more general timeframe bias is discussed. (shrink)

How do humans discover causal relations when the effect is not immediately observable? Previous experiments have uniformly demonstrated detrimental effects of outcome delays on causal induction. These findings seem to conflict with everyday causal cognition, where humans can apparently identify long-term causal relations with relative ease. Three experiments investigated whether the influence of delay on adult human causal judgements is mediated by experimentally induced assumptions about the timeframe of the causal relation in question, as suggested by Einhorn and Hogarth (1986). (...) Causal judgements generally decreased when a delay separated cause and effect. This decrease was less pronounced when the thematic context of the causal relation induced participants to expect a delay. Experiment 3 ruled out an alternative explanation of the effect based on variations of cue and outcome saliencies, and showed that detrimental effects of delay are reduced even more when instructions explicitly mentioned the timeframe of the causal relation in question. Knowledge thus mediates the impact of delay on human causal judgement. Implications for contemporary theories of human causal induction are discussed. (shrink)

Distinctions have been proposed between systems of reasoning for centuries. This article distills properties shared by many of these distinctions and characterizes the resulting systems in light of recent findings and theoretical developments. One system is associative because its computations reflect similarity structure and relations of temporal contiguity. The other is "rule based" because it operates on symbolic structures that have logical content and variables and because its computations have the properties that are normally assigned to rules. The systems serve (...) complementary functions and can simultaneously generate different solutions to a reasoning problem. The rule-based system can suppress the associative system but not completely inhibit it. The article reviews evidence in favor of the distinction and its characterization. (shrink)

Both philosophers and psychologists have argued for the existence of distinct kinds of explanations, including teleological explanations that cite functions or goals, and mechanistic explanations that cite causal mechanisms. Theories of causation, in contrast, have generally been unitary, with dominant theories focusing either on counterfactual dependence or on physical connections. This paper argues that both approaches to causation are psychologically real, with different modes of explanation promoting judgments more or less consistent with each approach. Two sets of experiments isolate the (...) contributions of counterfactual dependence and physical connections in causal ascriptions involving events with people, artifacts, or biological traits, and manipulate whether the events are construed teleologically or mechanistically. The findings suggest that when events are construed teleologically, causal ascriptions are sensitive to counterfactual dependence and relatively insensitive to the presence of physical connections, but when events are construed mechanistically, causal ascriptions are sensitive to both counterfactual dependence and physical connections. The conclusion introduces an account of causation, an "exportable dependence theory," that provides a way to understand the contributions of physical connections and teleology in terms of the functions of causal ascriptions. (shrink)

Causal queries about singular cases, which inquire whether specific events were causally connected, are prevalent in daily life and important in professional disciplines such as the law, medicine, or engineering. Because causal links cannot be directly observed, singular causation judgments require an assessment of whether a co‐occurrence of two events c and e was causal or simply coincidental. How can this decision be made? Building on previous work by Cheng and Novick (2005) and Stephan and Waldmann (2018), we propose a (...) computational model that combines information about the causal strengths of the potential causes with information about their temporal relations to derive answers to singular causation queries. The relative causal strengths of the potential cause factors are relevant because weak causes are more likely to fail to generate effects than strong causes. But even a strong cause factor does not necessarily need to be causal in a singular case because it could have been preempted by an alternative cause. We here show how information about causal strength and about two different temporal parameters, the potential causes' onset times and their causal latencies, can be formalized and integrated into a computational account of singular causation. Four experiments are presented in which we tested the validity of the model. The results showed that people integrate the different types of information as predicted by the new model. (shrink)