Abstract: Causal reasoning is a component of most human cognitive functions. It relates to establishing the presence of causal relationships among events. When a causal relationship exists, there is a good reason to believe that event of one type (consider as the cause) are systematically related to events of another type (consider as the effects). Smoking causes lung cancer, here causal claims allow to make prediction and decisions. ‘smoking causes cancer’ is true if smoking makes you more likely to get cancer than not smoking.
As usual, in real life, we deal with these kinds of probabilities and likelihoods rather than certainties. The purpose of this topic is to analyze aspects of causation relevant for discussing causal reasoning in diagnostic context. The analysis will discuss different conceptions of causal reasoning in medical diagnosis, decision making in the medical domain, as diagnostic reasoning, where a physician diagnoses diseases from observed symptoms. Research on diagnostic reasoning has a long tradition in psychology. Much of the literature on judgment and decision making has focused on the conditions under which people utilize base rate information and make judgments in accordance with a simple statistical model, called Bayes’s rule. As well as description versus experience causal reasoning analysis done two experiment, First, common effect model and Second, common cause model.
Each experiment contrasted the description-experience, description-only, and experience-only learning conditions, which are necessary to predict the failure of cause model reasoning. Researchers covered different conceptions of causal-reasoning in medical diagnosis using verbal information, also described versus experienced causal learning scenarios. Also, they concluded that human understand cause and effect but non-human animals like rat and monkey, may or may not understand cause and effect. Animals may use information about cause and effect to improve decision making and make inference about past and future events but it is difficult to imagine the safety of patients in diagnostic causal reasoning. As well as in the case of diagnostic causal-reasoning, It is an important issue to investigate how people understand and utilize natural language expression of uncertainty, both theoretically and empirically.
Introduction:The aim of this paper is to discuss causal reasoning from a perspective of cognitive psychology. Causal reasoning is related to cognitive competencies, which is a psychological construct the can’t be observed but can be inferred from an individual’s behavior or performance on the content relevant task. Diagnostic reasoning is part of our human’s daily life. A doctor diagnoses diseases from observed symptoms. Similarly, a cognitive reasons diagnostically while doing any experiment if that did not yield as the expected outcome of the experiment. Diagnostic reasoning concern inference from observed effects to unobserved causes of these effects. Another type of diagnostic causal reasoning, which is more focused on single binary effect to a single binary cause, called elemental diagnostic reasoning.
This basic type of diagnostic reasoning seems quite simple compared with real-world scenarios, which involves a complex network of multiple causes and multiple effects. It raises a number of critical questions about how people should reason diagnostically (normative model) and how people, in fact, do reason diagnostically (descriptive model). Recent research has found that decision from description and decision from experience can conclude in a robust and predictable way, that is known as the description-experience gap. This gap refers to the robust finding that learning about uncertain choices via description or experience results in systematically different choices.
In this study, researchers examine search and choice processes in decisions from experience involving medical outcomes. As well as they compared the processes both to decisions from experience involving gambles and for the decision from description involving the same medical outcomes.Description versus Experience Causal-Reasoning: This research mainly focused on explaining away and screening off failures, where researchers compared the tasks in which causal scenarios were merely described vs experienced.
In past few decades causal Bayes net has merged as the theoretical tool of model complex causal reasoning, where researchers represented causal knowledge as set of variables that encode causes and effects and set of causal arrows. These causal arrows represent causal influences directed from causes to effects. Consider a example of common effect model with two causes generating joint effect. This model represent as, thebacteria and viruses are two independent causes for fever.
Consider another example common cause model with one cause generating two effects. This model represent as, that a virus causes two different symptoms. In some studies (Rehder & Mayrhofer, 2003) researchers shown that causal Bayes net captures central features of human causal reasoning. With causal Bayes net, people draw different inference with common cause and common effect models. This will reflect their sensitivity to causal direction. Latest studies (Rehder et al.
, 2017) on judgment and decision making have uncovered differences between causal reasoning based on described versus experienced scenarios. For this researchers conducted some experiments about a common effect model (Experiment 1) and common cause model (Experiment 2). Each experiment contrasted the description-experience, description-only and experience-only learning conditions, which they expected would moderate the predicted failures of causal reasoning model.Procedure of experiments: Participants studied several computer screen of information about the domain and then performed the inference test. On Initial screen they presenter a cover story and a description of the domain’s three variables and their two values. Fig.
1 Shows that description-only and description-experience participants also observed the computer screens that presented the two causal relationship. After observing computer screens, participants took a multiple choice test of their knowledge, during the test, participants could return to the computer screen, from where they studied. Participants: For both experiment 144 undergraduates of New York university participated. In each experience-only, description-only, description-experience was manipulated in between the subjects.
Additionally, each study used the two between subjects counter balancing factors described earlier and subjects were assigned randomly.Results: Initial analysis concluded no effects in either experiments of which domain subjects learned of which version of the data sheet was used. In later phase it show 95% errors in bars (Fig 2). Human diagnostic reasoning is an important topic in cognitive psychology since 1970s, but to the best of our knowledge its has never been investigated how people reason diagnostically with verbal information. In general this is also same situation for causal reasoning. Causal reasoning generally provides numerical information or displays subjects with frequency information in the form of independent events similar experiments researchers did for description versus experience causal reasoning. With respect to verbal information, some research on causal reasoning done for qualitative patterns of causal inference, which used qualitative verbal information for describing the relevant causal relation.
Some previous research on judgment and decision making with verbal information was either used within subject design to explore the consistency, while using different modes of presenting knowledge, or focused on potential mismatch between proposed guidelines for converting numerical values to verbal expression. In this latest research (Meder et al,. 2017), task was to make inferences based on verbal information whose numerical equivalents have been elicited in different study with different subjects.Diagnostic causal-reasoning with verbal information: For investigating how people make probabilistic causal inference researcher used sequential diagnostic reasoning task. In which observer were sequentially presented with evidence as symptoms of patients and they had to judge the relative probability of binary cause event by chemical X or Y. In this experiment participants received either verbal or numerical information and asked to make diagnostic judgment for different type of symptoms combinations. In each round, three effects were presented sequentially. After each round, participants provided a probability regarding the two possible cause.
For each time, researcher used numerical estimates associated with the verbal term. As well as during the experiment one researchers used four verbal expressions “infrequently”, “frequently”, “occasionally” and “almost always” to convey the strength of the relation between causes and effects. Experiment 1: This experiment has two phases: First, a learning phase, where participant learned the strength of individual causal relations. In the learning phase, subject’s task was to learn the strength of the individual relations in a round by round fashion. On each round, subjects were shown a substance along with a symptom and had to estimate how frequently the substance cause the symptom. Second, a diagnostic reasoning phase, where participant were sequentially presented with symptoms of different patients and they had to make a diagnostic judgment after each symptoms.
During the experiments researchers did some manipulation also like, inform the participant about the strength of the relation between causes and effects. For instance, in verbal condition participant learned that “X almost always causes A” , while in numerical condition participant learned that “X causes A in 88% of cases” The results demonstrate that subjects were capable of making accurate inferences (with respect to the probabilities derived from the standard Bayes model) when reasoning with verbal information. Experiment 2: Main goal of this experiment was to test the robustness of the findings of experiment 1. In this experiment researcher used same procedure as experiment 1.
The main difference was the researchers used 4 different verbal expression and numerical equivalents. As in experiment participant first had to learn the strength of the relation between cause and effects. Since main goal of this experiment was to contrast diagnostic inferences based on the verbal versus numerical information. Similar as experiment 1, after the diagnostic inference phase, participants were tested for whether they still remembered the symptoms, what they learned at the beginning of the experiment.
Researchers excluded all participants who could not correctly reproduce the strength of at least 7 of the 8 causal relation after the diagnostic reasoning phase. At last 23 of the 84 participants failed to meet the criteria and were excluded. Remaining 61 participants were valid of experiment, 32 for verbal condition and 29 for numerical condition.
Experiment 3: In this experiment researchers increased the complexity of the task by using unequal prior probabilities for the two cause events. While in both previous events probability was 0.5, in experiment 3, one cause had a prior probability of 0.
33 and the other had 0.67. Result of experiment 1 and experiment 2 had remarkable results between numerical and verbal condition and also high accuracy. Conclusion: Research on causal reasoning has a long tradition in psychology. Much of the literature on judgment and decision-making has focused on the conditions under which people utilize base rate information and make judgments in accordance with a simple statistical model, Bayes’s rule. In this report, I tried to related diagnostic reasoning from the perspective of causal inference under uncertainty.
The discussion on elemental diagnostic reasoning illustrates that there is not a single normative benchmark for diagnostic reasoning under uncertainty against which human behavior can be evaluated, but that different ideas exist about what may constitute an appropriate standard of rational inference from effect(s) to cause(s). The fundamental concepts seems to be that causal connection are at least lawlike, with the nature world they have some kind of uniformity or regularity. It is certainly because by observing some uniform pattern in the occurrence of events. The regular appearance of the effect follow it’s cause, that we expect that cause will be followed by the effect. Causal-Reasoning have relation with neuroscience also it reviews neuroscience evidence indicating that mental model for causal inference are implemented within lateral cortex.