对归纳推理贝叶斯模型的检验

作者用实验比较检验特征归纳的贝叶斯模型、关联相似性模型、基于特征的归纳推理模型和相似性覆盖模型这四种模型。以大学生为被试的实验结果表明：1）在强关联强度一致时,被试的归纳推理基本符合贝叶斯模型和关联相似性模型的预测,在弱关联强度一致时,不符合这四种模型预测;2）在关联强度不一致时,关联强度效应不符合贝叶斯模型的预测;3）在关联强度一致时,贝叶斯模型和关联相似性模型的预测结果是一致的,不能区分两个模型。实验结果较多地支持贝叶斯模型和关联相似性模型。     

EFFECT OF REINFORCER MAGNITUDE ON PERFORMANCE MAINTAINED BY PROGRESSIVE‐RATIO SCHEDULES

This experiment examined the relationship between reinforcer magnitude and quantitative measures of performance on progressive‐ratio schedules. Fifteen rats were trained under a progressive‐ratio schedule in seven phases of the experiment in which the volume of a 0.6‐M sucrose solution reinforcer was varied within the range 6–300 μl. Overall response rates in successive ratios conformed to a bitonic equation derived from Killeen's (1994) Mathematical Principles of Reinforcement. The “specific activation” parameter, a, which is presumed to reflect the incentive value of the reinforcer, was a monotonically increasing function of reinforcer volume; the “response time” parameter, δ, which defines the minimum response time, increased as a function of reinforcer volume; the “currency” parameter, b, which is presumed to reflect the coupling of responses to the reinforcer, declined as a function of volume. Running response rate (response rate calculated after exclusion of the postreinforcement pause) decayed monotonically as a function of ratio size; the index of curvature of this function increased as a function of reinforcer volume. Postreinforcement pause increased as a function of ratio size. Estimates of a derived from overall response rates and postreinforcement pauses showed a modest positive correlation across conditions and between animals. Implications of the results for the quantification of reinforcer value and for the use of progressive‐ratio schedules in behavioral neuroscience are discussed.     

以信息增益模式探讨在高维数据上的多类别证型分类

在慢性胃炎研究上决策树归纳法之使用已渐增加,以决策树归纳法同时表示信息增益以区别胃炎症状对证型分类之贡献,能更正确地区别慢性胃炎。而信息增益已广泛地用于评估两分分类,仅有很少报导有关多类别慢性胃炎分类,它需要探索多类别慢性胃炎分类之信息增益。本研究提出基于名目及次序样本类结果的决策树归纳法所做的多类别慢性胃炎分类,例如,不同慢性胃炎次型的症状样本,评估是以决策树归纳法与Friedman-Goldszmid,HGC,Cheng判别法之正确率作比较。显示平均正确率64.9%优于前三者且提高1.55%。     

A Centipede for Intransitive Preferrers

In the standard money pump, an agent with cyclical preferences can avoid exploitation if he shows foresight and solves his sequential decision problem using backward induction (BI). This way out is foreclosed in a modified money pump, which has been presented in Rabinowicz (2000). There, BI will lead the agent to behave in a self-defeating way. The present paper describes another sequential decision problem of this kind, the Centipede for an Intransitive Preferrer, which in some respects is even more striking than the modified pump. In the new problem, the BI reasoning that implies self-defeating behavior does not rest on the controversial robustness assumption concerning beliefs in one's future rationality. This strengthens the claim that foresight cannot save the intransitive preferrer from a self-defeating course of action.     

Response–reinforcer contiguity versus response-rate–reinforcer-rate covariance in rats' lever pressing: Support for a multiscale view

The multiscale molar view sees behavior as a flow, like a river, extended in time. Matching theory expresses the way activities compete for time. Relative time taken by any activity depends on relative induction. The present experiment tested matching theory applied to concurrent contingent and noncontingent food. As adjunctive activities that compete with operant activity, we recorded hopper head entries and presses on a lever near the food hopper that had no programmed consequences. Eight naïve rats were first exposed to a variable-time 60 s schedule, which across conditions was gradually transformed into a variable-interval 60 s schedule by increasing the proportion of food that was delivered contingent on pressing a lever far from the hopper. Another group of 4 rats that had been trained to press a lever near a food hopper were introduced in the second condition, in which one food delivery was contingent on far-lever pressing. We found induction following a power function to describe pressing on the far lever (operant activity). Matching theory combined with power-function induction also accounted for adjunctive activity. Results with single contingent food deliveries provided little support for the molecular view that behavior consists of discrete responses “strengthened” by immediately following reinforcers.     

Avoidance,induction, and the illusion of reinforcement

Environmental events that impact reproductive success may be called phylogenetically important events (PIEs). Some promote reproductive success, like mates and food; others threaten reproductive success, like predators and injury. Beneficial PIEs induce activities that enhance them, and detrimental PIEs induce activities that mitigate or avoid them. Free-operant avoidance relies on electric shock as a proxy for injury, a PIE. One theory takes avoidance behavior to be reinforced by its reducing shock rate. A more complete explanation is that avoidance both reduces shock rate and is induced by the PIEs it usually prevents. Shocks received act in concert with shock-rate reduction, in a feedback system. Four parametric data sets were analyzed to show that avoidance is induced by received shock rate according to power functions. Avoidance is not reinforced at all; avoidance is induced by its failures. Induction explains not only avoidance itself, but also phenomena unique to avoidance, like warmup and effects of unavoidable shock. Induction explains behavior more generally than reinforcement, because induction explains not only food-maintained operant and nonoperant behavior, but also shock-maintained behavior, including avoidance. Reinforcement fails to explain behavior when reinforcement is defined as strengthening by consequences. Induction erases the distinction between consequences and antecedents.     

P3b反映基于相似性的图形类别归纳决策过程(309)

先前关于归纳推理的研究没有探讨过图形类别归纳任务中决策过程的事件相关电位(ERPs)特征。该研究采用图形材料,通过记录和分析图形类别归纳推理任务的结论评估阶段的脑电信号,探讨了不同类别条件下图形归纳推理决策过程的ERP特征。实验结果显示,推理决策过程中,当前提为一些狗具有某种属性特征时,结论一致类别(狗类及马类)比结论不一致类别下(椅子类)的归纳推理诱发了更大的P3b的波幅; 而结论一致类别下,类别内(狗类)比类别外(马类)推理诱发更大的后部脑区的P3b的波幅,反映了相似性对类别归纳的贡献。该研究为理解类别归纳过程的决策阶段的动态加工过程提供了神经电生理学证据。研究结果进一步支持了类别归纳的相似性理论模型。     

Food rejection and the development of food category-based induction in 2–6 years old children

We studied children’s inductive inferences within the domain of food categories. There has so far been little research on inductive reasoning about food among children, despite the theoretical and practical importance of knowing what knowledge children bring to the table and how they use it. We tested the hypotheses that children’s food category-based induction performances and their food rejection are negatively correlated, and that these performances are influenced by the colour typicality of the food items. We recruited 126 children aged 2–6 years, and administered a category-based induction task. Participants were successively shown 8 sets of three pictures containing one target picture (a vegetable) and two test pictures (a vegetable dissimilar in colour to the target picture and a fruit similar in colour to the target picture). For each set, participants were told a novel property about the target picture and asked to generalise this property to one of the two test pictures. Additionally, the parents of each child filled out a questionnaire about his or her food rejection tendencies. Results on accuracy (i.e. if participants generalised the properties according to category membership, not perceptual similarity) provided the first empirical evidence in favour of a negative relationship between children’s food rejection and food category-based induction.     

Sad people are more accurate at face recognition than happy people

Mood has varied effects on cognitive performance including the accuracy of face recognition (Lundh & Ost, 1996). Three experiments are presented here that explored face recognition abilities in mood-induced participants. Experiment 1 demonstrated that happy-induced participants are less accurate and have a more conservative response bias than sad-induced participants in a face recognition task. Using a remember/know/guess procedure, Experiment 2 showed that sad-induced participants had more conscious recollections of faces than happy-induced participants. Additionally, sad-induced participants could recognise all faces accurately, whereas, happy- and neutral-induced participants recognised happy faces more accurately than sad faces. In Experiment 3, these effects were not observed when participants intentionally learnt the faces, rather than incidentally learnt the faces. It is suggested that happy-induced participants do not process faces as elaborately as sad-induced participants.     

The roles of the central executive and visuospatial storage in mental arithmetic: A comparison across strategies

Previous research has demonstrated that working memory plays an important role in arithmetic. Different arithmetical strategies rely on working memory to different extents—for example, verbal working memory has been found to be more important for procedural strategies, such as counting and decomposition, than for retrieval strategies. Surprisingly, given the close connection between spatial and mathematical skills, the role of visuospatial working memory has received less attention and is poorly understood. This study used a dual-task methodology to investigate the impact of a dynamic spatial n-back task (Experiment 1) and tasks loading the visuospatial sketchpad and central executive (Experiment 2) on adults' use of counting, decomposition, and direct retrieval strategies for addition. While Experiment 1 suggested that visuospatial working memory plays an important role in arithmetic, especially when counting, the results of Experiment 2 suggested this was primarily due to the domain-general executive demands of the n-back task. Taken together, these results suggest that maintaining visuospatial information in mind is required when adults solve addition arithmetic problems by any strategy but the role of domain-general executive resources is much greater than that of the visuospatial sketchpad.