记忆与推理：理解思维的新视角

需积分: 10 9 浏览量更新于2024-07-22 收藏 1.16MB PDF 举报

"《Reasoning as Memory-Psychology Press(2014)》这本书聚焦于记忆与推理之间的紧密联系，旨在通过整合认知心理学领域的发现、理论和范式，深化我们对推理的理解。书中包含九章内容，专门探讨基本记忆过程如何影响推理和决策。各章节涉及的核心主题包括语义记忆与推理的关系、专家记忆在推理中的作用、识别记忆与归纳推理、工作记忆与推理，以及元记忆在推理中的应用。本书不仅提供了思考问题的广泛视角，还引发了关于在排除了记忆差异后，推理领域还有多少待解释的问题的思考。对于研究推理、决策，以及关注各种认知过程中记忆角色的本科生、研究生和研究人员来说，这是一本极富价值的读物。" 《Reasoning as Memory》一书揭示了记忆在推理过程中的核心地位。首先，它探讨了语义记忆与推理的关联性，语义记忆是我们储存和理解一般知识的系统，这对逻辑推理至关重要，因为它允许我们根据已知的事实和概念进行推断。其次，书中分析了专家记忆在推理中的角色。专家在特定领域的深厚知识往往能增强他们的推理能力，因为他们能迅速提取相关的信息来解决问题或作出决策。此外，作者还讨论了识别记忆与归纳推理的关系。识别记忆涉及到我们对过去经验的确认，而归纳推理是通过对个别案例的观察得出一般规律的过程。这两者之间的互动可能影响我们如何从个别实例中抽象出普遍原则。接着，书中深入研究了工作记忆与推理的相互作用。工作记忆是我们的临时心理工作空间，它在处理复杂任务时，如推理和解决问题，起着至关重要的作用。工作记忆的容量限制可能直接影响我们的推理效率和准确性。最后，元记忆在推理中的应用也得到了讨论。元记忆是指我们对自己记忆能力的认知，它影响我们何时信任自己的记忆，何时寻求外部信息。在推理中，良好的元记忆能够帮助我们判断哪些信息是可靠的，从而做出更准确的判断。通过这些章节，读者可以了解到记忆不仅仅是一个孤立的心理过程，而是与推理和决策紧密相连，共同构成了人类认知的重要组成部分。本书为理解这些复杂的心理机制提供了一个综合的框架，并鼓励进一步探索推理的深层认知机制。

INTRODUCTION

Reasoning and memory: A case for integration

Valerie A. Thompson and Aidan Feeney

Memory is the cabinet of imagination, the treasury of reason, the registry of

conscience and the council-chamber of thought.

Giambattista Basile

Like many edited books, this one had its origins in a very simple idea: to examine

the extent to which we can further our understanding of reasoning by integrating

findings, theories, and paradigms from the field of memory. There is a growing

acknowledgment of the importance of integrating the study of reasoning with

other areas of cognitive psychology, but in many cases, the theorising about such

relationships remains vague and only loosely developed. In contrast, this volume

consists of nine chapters that make explicit links between basic memory process

and reasoning and decision making, addressing topics such as the relationship

between semantic memory and reasoning, the role of expert memory in reasoning,

recognition memory and induction, working memory and reasoning, and

metamemory in reasoning. In addition to providing broad coverage of the

contribution of memory to thinking, the chapters also cover the field of thinking,

broadly writ, including deductive reasoning, inductive reasoning, expert reasoning,

representation in reasoning, hypothetical thinking, intuitive versus analytic

thinking, and problem solving. Ideally, anyone interested in thinking, reasoning or

decision making ought to find this volume a go-to resource that illustrates how

understanding basic memory processes can elucidate and advance theory in their

own area of work.

Late in the editing process, it became clear that this lacuna exists in both the

popular and the scientific press. When we tried to search for aphorisms or quotations

with which to start this preface, we found what appeared to be an endless supply

of elegantly expressed views on reason’s relationship to emotion; however, there

were relatively few on the relationship between reasoning and memory. This was

2 Introduction

all the more puzzling given that reasoning is often characterised as the process by

which we integrate remembered information with newly acquired information, by

which we draw out the implications of current knowledge, apply existing

knowledge to novel problems, challenge existing assumptions, and make changes

to our knowledge base. Reasoning requires short-term memory in which to

operate, and it also shares component processes with many types of memory. To

take just two examples: Dougherty and colleagues (Dougherty, Thomas & Lange,

2010; Lange et al., this volume) have shown how a global memory model can be

adapted to explain probability judgment and decision making. In the field of

artificial intelligence, Hélie and Sun (2010) have developed a cognitive architecture

(CLARION) based on fundamental principles of memory that can account for a

number of key reasoning phenomena.

To be sure, the reasoning literature has long acknowledged the central role of

working memory in reasoning. For example, we know that individual differences

in cognitive capacity, as measured either by IQ or proxies to IQ (e.g., Stanovich,

2011) or Working Memory Capacity (WMC) (e.g., De Neys 2006a, 2006b) are

strong predictors of performance in a wide range of reasoning tasks. This relationship

is a critical piece of evidence in support of Dual Process Theories of Reasoning

(Evans & Stanovich, 2013), which posit that WMC is necessary to engage higher-

order, analytic thinking. Indeed, Evans and Stanovich (2013) have argued that the

engagement of WMC is the defining feature of analytic thinking.

Thus, it is not surprising that several of the chapters in this book address the

relationship between WM and reasoning. Of interest, however, is that these

chapters illustrate how much has been learned about that relationship and how

sophisticated theorising in that domain has become. As a point of contrast, an

earlier book in this series was Logie and Gilhooly’s (1998) book Working Memory

and Thinking. Comparison between that book and this one is intriguing. As their

title suggests, Logie and Gilhooly were primarily concerned with one particular

aspect of memory, and the chapters in their book considered questions about

strategic use of working memory sub-systems in reasoning, how working memory

is involved in age differences in reasoning, and the role of working memory in

expertise and problem solving. One very telling fact is that whilst most of the

chapters in the earlier book referred to Baddeley’s multi-store model, only one of

the chapters in this book does. Another telling comparison is between the

contributions made by Graeme Halford, who is the only contributor common to

both books; Graeme has co-authored a chapter for this volume, and was the author

on a chapter in the earlier book. Reading both chapters side by side, one can

clearly see how thinking about working memory and its role in reasoning has

developed since the late 1990s. Whereas Halford (1998) was concerned with ideas

about the development of processing capacity and relational complexity, Halford,

Andrews and Wilson (this volume) couch their arguments about working

memory and relational reasoning in terms of Dual Process Theories (see Evans,

2008) and contemporary ideas about relational binding in working memory (see

Oberauer, Süß, Wilhelm & Sander, 2007). The comparison between the two

Introduction 3

chapters makes it clear that ideas about the relationship between working memory

and reasoning have developed in parallel to the development of ideas about

working memory. Interestingly, the chapter by Wilhelms and colleagues offers an

alternative view of the relationship between general cognitive resources and

reasoning, which would have been almost unthinkable in 1998.

If working memory provides a source of continuity between our book and

Logie and Gilhooly’s, our volume nicely illustrates that there are now multiple

relations between the literatures on memory and reasoning. One relation is in

terms of computational modelling, and two of the chapters in this book (Lange et

al.; Feeney et al.) illustrate how computational models of memory may be applied

to the study of hypothetical and inductive reasoning. One very attractive feature of

a modelling approach is that it requires one to be very specific about the

commonalities and differences. In the case of the Feeney et al. chapter, some of

the commonalities are surprising. Another relation between memory and reasoning

that has recently emerged is the centrality of ideas about metacognition. The

chapters by Topolinski and Ackerman and Thompson illustrate the importance

of reasoners’ monitoring of their own reasoning to how they regulate their

reasoning processes. This is a very new development in the study of reasoning,

which illustrates how theoretical advances in the study of memory may lead to

analogous advances in the study of reasoning.

In sum, this volume represents continuity and progress in the study of relations

between memory and reasoning. We hope that it will be read by reasoning and

memory researchers alike, and whilst the questions which have inspired all of the

work described here are psychological in nature, we expect cognitive scientists and

philosophers to find material to interest them. Logie and Gilhooly started their

book by bemoaning the “divide-and-rule” approach that had until that point been

taken by cognitive psychologists to the areas of memory and thinking. As the

chapters themselves – and the overview below – demonstrate, many more

researchers have now considered memory and reasoning side by side: we reason

about relations between novel events and things that are important to us;

information cued from the environment is often the starting point of hypothesising

about an event; we require space in working memory to carry out our reasoning

and to integrate the content of current thought with things that we already know,

etc. Stated in these terms, one wonders why reasoning researchers would ever have

imagined that it could be useful to study reasoning without reference to memory.

We look forward to the time when elegant phrases about the relationship between

memory and reasoning are easy to find.

Overview of the chapters

A number of different mechanisms have been proposed to mediate the relationship

between general cognitive capacity and reasoning: that WMC is needed for a

directed search of counter-examples from long-term memory (Markovits, Doyon,

& Simoneau, 2002; De Neys, Schaeken & d’Ydewalle, 2005a), that WMC limits

4 Introduction

the complexity of the representation that can be constructed from the problem

information (Johnson-Laird & Byrne, 2002), or that WM capacity is necessary to

overturn a pre-potent response and to simulate an alternative answer (Stanovich,

2011). Given the critical position accorded to WM capacity in almost all theories

of reasoning, it should not be surprising that four of our nine chapters address the

relationship between WM and reasoning.

In the first chapter, Nash Unsworth examines the issue from the perspective

of a memory researcher, and investigates the mechanisms of working memory that

support the relationship between WMC and reasoning. In this chapter, he re-

analyses data from a number of published studies involving many hundreds of

participants. He obtains a counter-intuitive finding, namely that the relationship

between WMC and reasoning is not mediated by the difficulty of the reasoning

items, and that WMC is equally implicated in solving easy and difficult items.

Instead, a detailed analysis of the components of working memory reveals that

individual differences in the scope of attention, attentional control, secondary

memory and interference control all contribute to the relationship between WM

and reasoning, and that as a collective, they are sufficient to account for the

correlation between WMC and reasoning.

The mechanisms identified by Nash resonate with many of those proposed in

reasoning theories: secondary memory is involved in the search for counter-

examples; resistance to interference is necessary to put aside a pre-potent response

as well as to resist using counter-examples that are inappropriate to the task at hand

(De Neys, Schaeken & d’Ydewalle, 2005b); differences in the scope of attention

will determine the complexity of the representation that one is able to maintain

and the success of simulating an alternative representation of a problem. Importantly,

however, Unsworth has shown that none of these abilities, in isolation, explains

the relationship between WMC and reasoning performance, a finding that should

encourage reasoning researchers to adopt a more multifaceted approach to

measuring WM. In addition, these findings need to be reconciled with others

(Stanovich & West, 2008) showing that capacity (measured as SAT scores) does

not correlate with performance on all reasoning tests. Perhaps such correlations

would emerge by adopting a broader operational definition of capacity.

In Chapter 2 Halford, Andrews, and Wilson develop a theory that addresses

the role of relational processes in reasoning and working memory. As with Dual

Process theorists, these authors argue that analytic thinking is WM-dependent.

Specifically, they argue that all forms of higher cognition depend on processing

relations in WM. These relations are used to formulate a model or representation

of the premises, which must be represented in WM. In this view, WM is a system

that enables the dynamic binding of information to a coordinate system (as per

Oberauer et al., 2007), such that reasoning involves the construction of new

relational representations in WM. Consequently, limitations to our ability to reason

are determined by the complexity of the relations that are bound in WM.

Henry Markovits, Chapter 3, also argues that the key to understanding reasoning

is to understand how information is represented in WM. He demonstrates that

Introduction 5

variability in conditional inferences is a function of the on-line retrieval of counter-

examples from semantic memory. This information, which is cued by the premises,

is incorporated into a mental model, which can be processed either by a WM-

intensive counter-example strategy or by an intuitive probabilistic strategy. Here, the

focus is on the interplay between the information presented to the reasoner, an active

retrieval process and representations in WM.

The focus of Chapter 4 is also on the role that cued retrieval plays in reasoning

outcomes. Lange, Buttaccio, Sprenger, Harbison, Thomas, and Dougherty

describe a computational model of hypothesis generation (HyGene) in which

memory plays the pivotal role. HyGene supposes that hypotheses are cued by

information in the environment and that the number of hypotheses that can be

actively represented is constrained by WM. This observation is important for a

variety of reasoning phenomena, given that many reasoning biases are thought to

occur because people represent only a single model of the situation at a time

(Evans, 2006). Using a combination of modelling and experimental data, the

authors show how HyGene accounts for a number of reasoning phenomena

associated with hypothesis generation and testing, including the circumstances

under which positive testing is preferred to diagnostic testing.

Whereas the focus of the first four chapters is on the limiting role that WMC

plays on reasoning, Wilhelms, Corbin, and Reyna, Chapter 5, outline an

alternative view. Fuzzy trace theory assumes that deliberate, analytic reasoning

requires access to rote, verbatim traces and that intuitions arise from gist-based

traces that encode meaning. Wilhelms et al. argue that reliance on gist processes

gives rise to both reasoning biases and expert knowledge. Because both gist and

verbatim processes are proposed to develop with age and experience, surprising

developmental reversals are observed. For example, gist-based representations are

posited to be the source of framing effects, which get stronger as people progress

from childhood to adolescence. Similarly, expert reasoners (e.g., cardiologists) are

posited to rely on gist-based processing, which allows them to make accurate

categorical assessments of risk, but none the less be subject to reasoning fallacies,

such as the disjunction effect.

The next two chapters also focus on the role of the knowledge structures that

support reasoning. Feeney, Hayes, and Heit (Chapter 6) explore the

proposition that many of the cognitive mechanisms that support recognition

memory also underlie category-based induction. For example, the authors argue

that it is possible to make inferences about the processes required to make

inductive judgments from how well people are subsequently able to recognise

the targets of that reasoning. The “inference then recognition” paradigm asks

participants to reason about novel or familiar categories. This is followed by a

surprise recognition test, which, for example, can help to distinguish between

inferences that are made in a category-based way and those that are made on the

basis of featural overlap. They go on to argue that a computational model of

recognition memory captures many important features of the people’s judgments

about the strength of inductive arguments. That is, judgments of the strength of

剩余202页未读，继续阅读

Jerymouse

粉丝: 2
资源: 7

记忆与推理：理解思维的新视角

Bayesian-Reasoning-and-Machine-Learning

SHL-Verbal-Reasoning-Test1-Solutions

Probabilistic Reasoning in Multi-Agent Systems

Causal Reasoning from Meta-reinforcement Learning的主要方法

Demonstrate that you can generate the series for ex as (define exp-series (stream-cons 1 (integrate-series-tail exp-series))) Explain the reasoning behind this definition. Show how to generate the series for sine and cosine, in a similar way, as a pair of mutually recursive definitions.

Some of the most difficult problems that AI has to solve

Fact-Checking Complex Claims with Program-Guided Reasoning的代码复现

production reasoning在知识图谱中是啥意思

Bayesian Reasoning and Machine Learning第二章代码

最新资源