Since its conception nearly 20 years ago, Logic Programming - the idea of using logic as a programming language - has been developed to the point where it now plays an important role in areas such as database theory, artificial intelligence and software engineering. However, there are still many challenging research issues to be addressed and the UK branch of the Association for Logic Programming was set up to provide a forum where the flourishing research community could discuss important issues (...) of Logic Programming which were often by-passed at the large international conferences. This volume contains the twelve papers which were presented at the ALPUK's 3rd conference which was held in Edinburgh, 10-12 April 1991. The aim of the conference was to give a broad but detailed technical insight into the work currently being done in this field, both in the UK and by researchers as far afield as Canada and Bulgaria. The breadth of interest in this area of Computer Science is reflected in the range of the papers which cover - amongst other areas - massively parallel implementation, constraint logic programming, circuit modelling, algebraic proof of program properties, deductive databases, specialised editors and standardisation. The resulting volume gives a good overview of the current progress being made in the field and will be of interest to researchers and students of any aspects of logic programming, parallel computing or database techniques and management. (shrink)

states are to be understood in terms of their functional relationships to other mental states, not in terms of their material instantiation in any particular kind of hardware. But the argument that material instantiation is irrelevant to functional..

Internet of Things is expanding and evolves into all aspects of the society. Research and developments in the field of IoT have shown the possibility of producing huge volume of data and computation among different devices of the IoT. The data collected from IoT devices are transferred to a central server which can further be retrieved and accessed by the service providers for analyzing, processing, and using. Industrial Internet of Health Things is the expansion of the Internet of Health Things (...) which plays an important role in observing, consulting, monitoring, and treatment process of remote exchange data processes. The linkage of computation and interoperability are supported through various intelligent sensors, controllers, and actuators. The role of parallel computing for efficient and Intelligent Industrial Internet of Health Things is obvious to analyze and process different healthcare situations. A detailed overview of this existing literature is needed through which the research community will provide new solutions for efficient healthcare with the help of IoT based on parallel computing. Therefore, the current study presents a detailed overview of the existing literature for facilitating IIoHT. (shrink)

Wilfried Sieg and John Byrnes. AnModel for Parallel Computation: Gandy's Thesis.

This article discusses the properties of a controllable, flexible, hybrid parallel computing architecture that potentially merges pattern recognition and arithmetic. Humans perform integer arithmetic in a fundamentally different way than logic-based computers. Even though the human approach to arithmetic is both slow and inaccurate it can have substantial advantages when useful approximations are more valuable than high precision. Such a computational strategy may be particularly useful when computers based on nanocomponents become feasible because it offers a way (...) to make use of the potential power of these massively parallel systems. Because the system architecture is inspired by neuroscience and is applied to cognitive problems, occasional mention is made of experimental data from both fields when appropriate. (shrink)

I recently argued that the position in the philosophy of mind called functionalism is undermined by the importance of recent work on parallel computation (Thagard, 1986). In reply, Krellenstein (1987) contends that parallelism does not have the philosophical significance I claimed for it. Although his contentions are plausible if one focuses on what is in principle computationally possible, they fail if one looks at real problems.

I recently argued that the position in the philosophy of mind called functionalism is undermined by the importance of recent work on parallel computation (Thagard, 1986). In reply, Krellenstein (1987) contends that parallelism does not have the philosophical significance I claimed for it. Although his contentions are plausible if one focuses on what is in principle computationally possible, they fail if one looks at real problems.

The purpose of this paper is to show that there exist star-finite tree-structured sets in which the computations of parallel programs can be faithfully embedded, and that the theory of star-finite sets and relations therefore provides a new tool for the analysis of non-deterministic computations.

In his classic paper On Computable Numbers Turing analyzed what can be done by a human computor in a routine, “mechanical” way. He argued that mechanical op-erations obey locality conditions and are carried out on configurations satisfying boundedness conditions. Processes meeting these restrictive conditions can be shown to be computable by a Turing machine. Turing viewed memory limitations of computors as the ultimate reason for the restrictive conditions. In contrast, Gandy analyzed in his paper Church’s Thesis and Principles for Mechanisms (...) what can be done by “discrete deterministic mechanical devices” and appealed to physical considerations to motivate restrictive principles. These devices include, crucially, ones that carry out parallel processes, e.g., cellular automata. Following Turing’s methodological ways and guided by Conway’s “Game of Life” as a paradigm, Gandy formulated four restrictive principles and proved that devices satisfying them are computationally equivalent to Turing machines. (shrink)

This article discusses the properties of a controllable, flexible, hybrid parallel computing architecture that potentially merges pattern recognition and arithmetic. Humans perform integer arithmetic in a fundamentally different way than logic-based computers. Even though the human approach to arithmetic is both slow and inaccurate it can have substantial advantages when useful approximations ( intuition ) are more valuable than high precision. Such a computational strategy may be particularly useful when computers based on nanocomponents become feasible because it (...) offers a way to make use of the potential power of these massively parallel systems. Because the system architecture is inspired by neuroscience and is applied to cognitive problems, occasional mention is made of experimental data from both fields when appropriate. (shrink)

Andrew Boucher (1997) argues that ``parallel computation is fundamentally different from sequential computation'' (p. 543), and that this fact provides reason to be skeptical about whether AI can produce a genuinely intelligent machine. But parallelism, as I prove herein, is irrelevant. What Boucher has inadvertently glimpsed is one small part of a mathematical tapestry portraying the simple but undeniable fact that physical computation can be fundamentally different from ordinary, ``textbook'' computation (whether parallel or sequential). This tapestry does indeed (...) immediately imply that human cognition may be uncomputable. (shrink)

This paper shows that the basic logic induced by the parallel recurrence $\hspace {-2pt}\mbox {\raisebox {-0.01pt}{\@setfontsize \small {7}{8}$\wedge$}\hspace {-3.55pt}\raisebox {4.5pt}{\tiny $\mid$}\hspace {2pt}}$ of computability logic (i.e., the one in the signature $\{\neg,$\wedge$,\vee,\hspace {-2pt}\mbox {\raisebox {-0.01pt}{\@setfontsize \small {7}{8}$\wedge$}\hspace {-3.55pt}\raisebox {4.5pt}{\tiny $\mid$}\hspace {2pt}},\hspace {-2pt}\mbox {\raisebox {0.12cm}{\@setfontsize \small {7}{8}$\vee$}\hspace {-3.6pt}\raisebox {0.02cm}{\tiny $\mid$}\hspace {2pt}}\}$ ) is a proper superset of the basic logic induced by the branching recurrence $\mbox {\raisebox {-0.05cm}{$\circ$}\hspace {-0.11cm}\raisebox {3.1pt}{\tiny $\mid$}\hspace {2pt}}$ (i.e., the one in the signature $\{\neg,$\wedge$,\vee,\mbox {\raisebox {-0.05cm}{$\circ$}\hspace (...) {-0.11cm}\raisebox {3.1pt}{\tiny $\mid$}\hspace {2pt}},\mbox {\raisebox {0.12cm}{$\circ$}\hspace {-0.115cm}\raisebox {0.02cm}{\tiny $\mid$}\hspace {2pt}}\}$ ). The latter is known to be precisely captured by the cirquent calculus system CL15 , conjectured by Japaridze to remain sound—but not complete—with $\hspace {-2pt}\mbox {\raisebox {-0.01pt}{\@setfontsize \small {7}{8}$\wedge$}\hspace {-3.55pt}\raisebox {4.5pt}{\tiny $\mid$}\hspace {2pt}}$ instead of $\mbox {\raisebox {-0.05cm}{$\circ$}\hspace {-0.11cm}\raisebox {3.1pt}{\tiny $\mid$}\hspace {2pt}}$ . The present result is obtained by positively verifying that conjecture. A secondary result of the paper is showing that $\hspace {-2pt}\mbox {\raisebox {-0.01pt}{\@setfontsize \small {7}{8}$\wedge$}\hspace {-3.55pt}\raisebox {4.5pt}{\tiny $\mid$}\hspace {2pt}}$ is strictly weaker than $\mbox {\raisebox {-0.05cm}{$\circ$}\hspace {-0.11cm}\raisebox {3.1pt}{\tiny $\mid$}\hspace {2pt}}$ in the sense that, while $\mbox {\raisebox {-0.05cm}{$\circ$}\hspace {-0.11cm}\raisebox {3.1pt}{\tiny $\mid$}\hspace {2pt}}F$ logically implies $\hspace {-2pt}\mbox {\raisebox {-0.01pt}{\@setfontsize \small {7}{8}$\wedge$}\hspace {-3.55pt}\raisebox {4.5pt}{\tiny $\mid$}\hspace {2pt}}F$ , the reverse does not hold. (shrink)

In a recent paper, Lyngzeidetson [1990] has claimed that a type of parallel computer called the ‘Connection Machine’ instantiates architectural principles which will ‘revolutionize which "functions" of the human mind can and cannot be modelled by (non-human) computational automata.’ In particular, he claims that the Connection Machine architecture shows the anti-mechanist argument from Gödel's theorem to be false for at least one kind of parallel computer. In the first part of this paper, I argue that Lyngzeidetson's claims are (...) not supported by his arguments; in the second part I consider some other aspects of parallel computation which may be of theoretical significance in cognitive science. (shrink)

The efficient processing of large-scale data has very important practical value. In this study, a data mining platform based on Hadoop distributed file system was designed, and then K-means algorithm was improved with the idea of max-min distance. On Hadoop distributed file system platform, the parallelization was realized by MapReduce. Finally, the data processing effect of the algorithm was analyzed with Iris data set. The results showed that the parallel algorithm divided more correct samples than the traditional algorithm; in (...) the single-machine environment, the parallel algorithm ran longer; in the face of large data sets, the traditional algorithm had insufficient memory, but the parallel algorithm completed the calculation task; the acceleration ratio of the parallel algorithm was raised with the expansion of cluster size and data set size, showing a good parallel effect. The experimental results verifies the reliability of parallel algorithm in big data processing, which makes some contributions to further improve the efficiency of data mining. (shrink)

One way of coping with agreement of features in French is to perform two parallel computations, one in the free pregroup of syntactic types, the other in that of feature types. Technically speaking, this amounts to working in the direct product of two free pregroups.

The cerebral cortex is a rich and diverse structure that is the basis of intelligent behavior. One of the deepest mysteries of the function of cortex is that neural processing times are only about one hundred times as fast as the fastest response times for complex behavior. At the very least, this would seem to indicate that the cortex does massive amounts of parallel computation.This paper explores the hypothesis that an important part of the cortex can be modeled as (...) a connectionist computer that is especially suited for parallel problem solving. The connectionist computer uses a special representation, termed value unit encoding, that represents small subsets of parameters in a way that allows parallel access to many different parameter values. This computer can be thought of as computing hierarchies of sensorimotor invariants. The neural substrate can be interpreted as a commitment to data structures and algorithms that compute invariants fast enough to explain the behavioral response times. A detailed consideration of this model has several implications for the underlying anatomy and physiology. (shrink)

Although several theories of online syntactic processing assume the parallel activation of multiple syntactic representations, evidence supporting simultaneous activation has been inconclusive. Here, the continuous and non‐ballistic properties of computer mouse movements are exploited, by recording their streaming x, y coordinates to procure evidence regarding parallel versus serial processing. Participants heard structurally ambiguous sentences while viewing scenes with properties either supporting or not supporting the difficult modifier interpretation. The curvatures of the elicited trajectories revealed both an effect of (...) visual context and graded competition between simultaneously active syntactic representations. The results are discussed in the context of 3 major groups of theories within the domain of sentence processing. (shrink)

By Parallel Reasoning is the first comprehensive philosophical examination of analogical reasoning in more than forty years designed to formulate and justify standards for the critical evaluation of analogical arguments. It proposes a normative theory with special focus on the use of analogies in mathematics and science. In recent decades, research on analogy has been dominated by computational theories whose objective has been to model analogical reasoning as a psychological process. These theories have devoted little attention to normative questions. (...) In this book Bartha proposes that a good analogical argument must articulate a clear relationship that is capable of generalization. This idea leads to a set of distinct models for the critical analysis of prominent forms of analogical argument. The same core principle makes it possible to relate analogical reasoning to norms and values of scientific practice. Reasoning by analogy is justified because it strikes an optimal balance between conservative values, such as simplicity and coherence, and progressive values, such as fruitfulness and theoretical unification. Analogical arguments are also justified by appeal to symmetry--like cases are to be treated alike. In elaborating the connection between analogy and these broad epistemic principles, By Parallel Reasoning offers a novel contribution to explaining how analogies can play an important role in the confirmation of scientific hypotheses. (shrink)

Gualtiero Piccinini articulates and defends a mechanistic account of concrete, or physical, computation. A physical system is a computing system just in case it is a mechanism one of whose functions is to manipulate vehicles based solely on differences between different portions of the vehicles according to a rule defined over the vehicles. Physical Computation discusses previous accounts of computation and argues that the mechanistic account is better. Many kinds of computation are explicated, such as digital vs. analog, serial vs. (...) parallel, neural network computation, program-controlled computation, and more. Piccinini argues that computation does not entail representation or information processing although information processing entails computation. Pancomputationalism, according to which every physical system is computational, is rejected. A modest version of the physical Church-Turing thesis, according to which any function that is physically computable is computable by Turing machines, is defended. (shrink)

What is the mind? How does it work? How does it influence behavior? Some psychologists hope to answer such questions in terms of concepts drawn from computer science and artificial intelligence. They test their theories by modeling mental processes in computers. This book shows how computer models are used to study many psychological phenomena--including vision, language, reasoning, and learning. It also shows that computer modeling involves differing theoretical approaches. Computational psychologists disagree about some basic questions. For instance, should the (...) mind be modeled by digital computers, or by parallel-processing systems more like brains? Do computer programs consist of meaningless patterns, or do they embody (and explain) genuine meaning? (shrink)

Computational methods such as computer simulations, Monte Carlo methods, and agent-based modeling have become the dominant techniques in many areas of science. Extending Ourselves contains the first systematic philosophical account of these new methods, and how they require a different approach to scientific method. Paul Humphreys draws a parallel between the ways in which such computational methods have enhanced our abilities to mathematically model the world, and the more familiar ways in which scientific instruments have expanded our access to (...) the empirical world. This expansion forms the basis for a new kind of empiricism, better suited to the needs of science than the older anthropocentric forms of empiricism. Human abilities are no longer the ultimate standard of epistemological correctness.Humphreys also includes arguments for the primacy of properties rather than objects, the need to consider technological constraints when appraising scientific methods, and a detailed account of how the path from computational template to scientific application is constructed. This last feature allows us to hold a form of selective realism in which anti-realist arguments based on formal reconstructions of theories can be avoided. One important consequence of the rise of computational methods is that the traditional organization of the sciences is being replaced by an organization founded on computational templates.Extending Ourselves will be of interest to philosophers of science, epistemologists, and to anyone interested in the role played by computers in modern science. (shrink)

In quantum computation non classical features such as superposition states and entanglement are used to solve problems in new ways, impossible on classical digital computers.We illustrate by Deutsch algorithm how a quantum computer can use superposition states to outperform any classical computer. We comment on the view of a quantum computer as a massive parallel computer and recall Amdahls law for a classical parallel computer. We argue that the view on quantum computation as a massive parallel (...) computation disregards the presence of entanglement in a general quantum computation and the non classical way in which parallel results are combined to obtain the final output. (shrink)

Because it is time-dependent, parallel computation is fundamentally different from sequential computation. Parallel programs are non-deterministic and are not effective procedures. Given the brain operates in parallel, this casts doubt on AI's attempt to make sequential computers intelligent.

In recent years quantum probability models have been used to explain many aspects of human decision making, and as such quantum models have been considered a viable alternative to Bayesian models based on classical probability. One criticism that is often leveled at both kinds of models is that they lack a clear interpretation in terms of psychological mechanisms. In this paper we discuss the mechanistic underpinnings of a quantum walk model of human decision making and response time. The quantum walk (...) model is compared to standard sequential sampling models, and the architectural assumptions of both are considered. In particular, we show that the quantum model has a natural interpretation in terms of a cognitive architecture that is both massively parallel and involves both co-operative and competitive interactions between units. Additionally, we introduce a family of models that includes aspects of the classical and quantum walk models. (shrink)

Proposals for quantum computation rely on superposed states implementing multiple computations simultaneously, in parallel, according to quantum linear superposition (e.g., Benioff, 1982; Feynman, 1986; Deutsch, 1985, Deutsch and Josza, 1992). In principle, quantum computation is capable of specific applications beyond the reach of classical computing (e.g., Shor, 1994). A number of technological systems aimed at realizing these proposals have been suggested and are being evaluated as possible substrates for quantum computers (e.g. trapped ions, electron spins, quantum dots, nuclear (...) spins, etc., see Table 1; Bennett, 1995; and Barenco, 1995). The main obstacle to realization of quantum computation is the problem of interfacing to the system (input, output) while also protecting the quantum state from environmental decoherence. If this problem can be overcome, then present day classical computers may evolve to quantum computers. (shrink)

Parallel processing as a method to improve computer performance has become a development trend. Based on rough set theory and divide-and-conquer idea of knowledge reduction, this paper proposes a classification method that supports parallel attribute reduction processing, the method makes the relative positive domain which needs to be calculated repeatedly independent, and the independent relative positive domain calculation could be processed in parallel; thus, attribute reduction could be handled in parallel based on this classification method. Finally, (...) the proposed algorithm and the traditional algorithm are analyzed and compared by experiments, and the results show that the proposed method in this paper has more advantages in time efficiency, which proves that the method could improve the processing efficiency of attribute reduction and makes it more suitable for massive data sets. (shrink)

ABSTRACT Parallel processing systems can carry out computational tasks which would be impossible to be carried out by sequential systems. Cognitive psychologists are discovering that brains do not operate on a sequential ordering of tasks, but along parallel processing models. Sequential ordering is abandoned in the new generation computers, which are being designed on evolving parallel processing models. My proposal consists in applying the parallel processing principles to the state, creating a ‘parallel governing’model for (...) the decision‐making procedures at the political level, in place of the present sequentially ordered procedures. I describe the main principles of current parallel processing models, and use them towards the creation of a parallel governing system. The most fundamental principle is the ‘De‐Centralisation Principle’, which requires that there be no centralised unit with special rights to information, or to policy‐, and decision‐making authority. In parallel governing, political and moral principles are built into the structure of the system, which consists of units of specialised interests and powers, with specialised channels of communication between them. I close by delineating the main differences between parallel governing and Nozick's utopia and Hayek's neo‐libertarianism. (shrink)

Context: At present, we lack a common understanding of both the process of cognition in living organisms and the construction of knowledge in embodied, embedded cognizing agents in general, including future artifactual cognitive agents under development, such as cognitive robots and softbots. Purpose: This paper aims to show how the info-computational approach (IC) can reinforce constructivist ideas about the nature of cognition and knowledge and, conversely, how constructivist insights (such as that the process of cognition is the process of life) (...) can inspire new models of computing. Method: The info-computational constructive framework is presented for the modeling of cognitive processes in cognizing agents. Parallels are drawn with other constructivist approaches to cognition and knowledge generation. We describe how cognition as a process of life itself functions based on info-computation and how the process of knowledge generation proceeds through interactions with the environment and among agents. Results: Cognition and knowledge generation in a cognizing agent is understood as interaction with the world (potential information), which by processes of natural computation becomes actual information. That actual information after integration becomes knowledge for the agent. Heinz von Foerster is identified as a precursor of natural computing, in particular bio computing. Implications: IC provides a framework for unified study of cognition in living organisms (from the simplest ones, such as bacteria, to the most complex ones) as well as in artifactual cognitive systems. Constructivist content: It supports the constructivist view that knowledge is actively constructed by cognizing agents and shared in a process of social cognition. IC argues that this process can be modeled as info-computation. (shrink)

I offer an explication of the notion of computer, grounded in the practices of computability theorists and computer scientists. I begin by explaining what distinguishes computers from calculators. Then, I offer a systematic taxonomy of kinds of computer, including hard-wired versus programmable, general-purpose versus special-purpose, analog versus digital, and serial versus parallel, giving explicit criteria for each kind. My account is mechanistic: which class a system belongs in, and which functions are computable by which system, depends on the (...) system's mechanistic properties. Finally, I briefly illustrate how my account sheds light on some issues in the history and philosophy of computing as well as the philosophy of mind. (shrink)

Distributed programming paradigms such as MapReduce and Spark have alleviated sequential bottleneck while mining of massive transaction databases. Of significant importance is mining High Utility Itemset that incorporates the revenue of the items purchased in a transaction. Although a few algorithms to mine HUIs in the distributed environment exist, workload skew and data transfer overhead due to shuffling operations remain major issues. In the current study, Parallel Utility Computation algorithm has been proposed with novel grouping and load balancing strategies (...) for an efficient mining of HUIs in a distributed environment. To group the items, Transaction Weighted Utility values as a degree of transaction similarity is employed. Subsequently, these groups are assigned to the nodes across the cluster by taking into account the mining load due to the items in the group. Experimental evaluation on real and synthetic datasets demonstrate that PUC with TWU grouping in conjunction with load balancing converges mining faster. Due to reduced data transfer, and load balancing-based assignment strategy, PUC outperforms different grouping strategies and random assignment of groups across the cluster. Also, PUC is shown to be faster than PHUI-Growth algorithm with a promising speedup. (shrink)

Shastri & Ajjanagadde argue convincingly that both structured connectionist networks and parallel dynamic inferencing are necessary for reflexive reasoning - a kind of inferencing and reasoning that occurs rapidly, spontaneously, and without conscious effort, and which seems necessary for everyday tasks such as natural language understanding. As S&A describe, reflexive reasoning requires a solution to thedynamic binding problem, that is, how to encode systematic and abstract knowledge and instantiate it in specific situations to draw appropriate inferences. Although symbolic artificial (...) intelligence systems trivially solve the dynamic binding problem using computers' registers and pointers, it has remained a difficult problem for connectionist systems. S&A's temporal synchrony solution to the dynamic binding problem using synchronous firing of argument units and the entities that are bound to them illustrates one way in which connectionist networks can do thisusing a constrained but important class of long-term knowledge rules. Their structured connectionist solution allows dynamic inferencing to proceed in parallel and therefore has a number of advantages for reflexive reasoning over most other connectionist and symbolic systems. (shrink)

This paper defines natural hierarchies of function and relation classes □i,kc and Δi,kc, constructed from parallel complexity classes in a manner analogous to the polynomial-time hierarchy. It is easily shown that □i−1,kp □c,kc □i,kp and similarly for the Δ classes. The class □i,3c coincides with the single-valued functions in Buss et al.'s class , and analogously for other growth rates. Furthermore, the class □i,kc comprises exactly the functions Σi,kb-definable in Ski−1, and if Tki−1 is Σi,kb-conservative over Ski−1, then □i,kp (...) is completely parallelizable. All functions in □i,kc are Σi,kb-definable in Rki; this suffices to show that if the known Σi,kb conservativity between R3i and S3i−1 extends to R2i and S2i−1, then NC = NC1 relative to an oracle in PH. We prove a KPT-style witnessing theorem for Ski using constantly many rounds of □i,kc interactive computation, and thus show that if Ski ≡ Rki+1 then the bounded arithmetic hierarchy collapses, provably in Ski. (shrink)

A computational modeling approach was used to test one possible explanation for the limited capacity of the subitizing phenomenon. Most existing models of this phenomenon associate the subitizing span with an assumed structural limitation of the human information processing system. In contrast, we show how this limit might emerge as the combinatorics of the space of enumeration problems interacts with the human cognitive architecture in the context of an enumeration task. Subitizing‐like behavior was generated in two different models of enumeration, (...) one based on the ACT‐R cognitive architecture and the other based on the principles of parallel distributed processing (PDP). Our results provide good qualitative fits to results obtained in a variety of empirical studies. (shrink)

Computer simulation modeling is an important part of contemporary scientific practice but has not yet received much attention from philosophers. The present project helps to fill this lacuna in the philosophical literature by addressing three questions that arise in the context of computer simulation of Earth's climate. Computer simulation experimentation commonly is viewed as a suspect methodology, in contrast to the trusted mainstay of material experimentation. Are the results of computer simulation experiments somehow deeply problematic in ways that the results (...) of material experiments are not? I argue against categorical skepticism toward the results of computer simulation experiments by revealing important parallels in the epistemologies of material and computer simulation experimentation. It has often been remarked that simple computer simulation models---but not complex ones---contribute substantially to our understanding of the atmosphere and climate system. Is this view of the relative contributions of simple and complex models tenable? I show that both simple and complex climate models can promote scientific understanding and argue that the apparent contribution of simple models depends upon whether a causal or deductive account of scientific understanding is adopted. When two incompatible scientific theories are under consideration, they typically are viewed as competitors, and we seek evidence that refutes at least one of the theories. In the study of climate change, however, logically incompatible computer simulation models are accepted as complementary resources for investigating future climate. How can we make sense of this use of incompatible models? I show that a collection of incompatible climate models persists in part because of difficulties faced in evaluating and comparing climate models. I then discuss the rationale for using these incompatible models together and argue that this climate model pluralism has both competitive and integrative components. (shrink)