Locked and unlocked strategies are at the center of this article, as ways of shedding new light on the cognitive aspects of deep learning machines. The character and the role of these cognitive strategies, which are occurring both in humans and in computational machines, is indeed strictly related to the generation of cognitive outputs, which range from weak to strong level of knowledge creativity. I maintain that these differences lead to important consequences when we analyze computational AI programs, such as (...) AlphaGo, which aim at performing various kinds of abductive hypothetical reasoning. In these cases, the programs are characterized by _locked_ abductive strategies: they deal with weak (even if sometimes amazing) kinds of hypothetical creative reasoning, because they are limited in what I call eco-cognitive openness, which instead qualifies human cognizers who are performing higher kinds of abductive creative reasoning, where cognitive strategies are instead _unlocked_. (shrink)

Herbert Simon’s work on scientific discovery deserves serious attention by philosophers of science for several reasons. First, Simon was an early advocate of rational scientific discovery, contra Popper and logical empiricist philosophers of science (Simon 1966). This proposal spurred on investigation of scientific discovery in philosophy of science, as philosophers used and developed Simon’s notions of “problem solving” and “heuristics” in attempts to provide rational accounts of scientific discovery (See Nickles 1980a, Wimsatt 1980). Second, Simon promoted and developed many of (...) the crucial techniques and methods used in cognitive science. One is the use of computers to model internal psychological processes, a technique central to his account of scientific discovery. Another is protocol analysis, the use of the verbal reports of experimental subjects in psychology to construct accounts of their psychological processes. Protocol analysis is given a detailed formulation by Simon (Simon and Ericsson 1984), and is modified for use in the study of scientific cognition in the paper on Krebs (Kulkarni and Simon 1988). (shrink)

Construction of a robot discoverer can be treated as the ultimate success of automated discovery. In order to build such an agent we must understand algorithmic details of the discovery processes and the representation of scientific knowledge needed to support the automation. To understand the discovery process we must build automated systems. This paper investigates the anatomy of a robot-discoverer, examining various components developed and refined to a various degree over two decades. We also clarify the notion of autonomy of (...) an artificial agent, and we discuss the ways in which machine discoverers become more autonomous. Finally we summarize the main principles useful in construction of automated discoverers and we discuss various possible limitations of automation. (shrink)

The term conceptual simulation refers to a type of everyday reasoning strategy commonly called “what if” reasoning. It has been suggested in a number of contexts that this type of reasoning plays an important role in scientific discovery; however, little direct evidence exists to support this claim. This article proposes that conceptual simulation is likely to be used in situations of informational uncertainty, and may be used to help scientists resolve that uncertainty. We conducted two studies to investigate the relationship (...) between conceptual simulation and informational uncertainty. Study 1 was an in vivo study of expert scientists; the results suggest that scientists do use conceptual simulation in situations of informational uncertainty, and that they use conceptual simulation to make inferences from their data using the analogical reasoning process of alignment by similarity detection. Study 2 experimentally manipulated experts' level of uncertainty and provides further support for the hypothesis that conceptual simulation is more likely to be used in situations of informational uncertainty. Finally, we discuss the relationship between conceptual simulation and other types of reasoning using qualitative mental models. (shrink)

We conducted two in vivo studies to explore how scientists respond to anomalies. Based on prior research, we identify three candidate strategies: mental simulation, mental manipulation of an image, and comparison between images. In Study 1, we compared experts in basic and applied domains (physics and meteorology). We found that the basic scientists used mental simulation to resolve an anomaly, whereas applied science practitioners mentally manipulated the image. In Study 2, we compared novice and expert meteorologists. We found that unlike (...) experts, novices used comparison to address anomalies. We discuss the nature of expertise in the two kinds of science, the relationship between the type of science and the task performed, and the relationship of the strategies investigated to scientific creativity. (shrink)

How can anyone be rational in a world where knowledge is limited, time is pressing, and deep thought is often an unattainable luxury? Traditional models of unbounded rationality and optimization in cognitive science, economics, and animal behavior have tended to view decision-makers as possessing supernatural powers of reason, limitless knowledge, and endless time. But understanding decisions in the real world requires a more psychologically plausible notion of bounded rationality. In Simple heuristics that make us smart (Gigerenzer et al. 1999), we (...) explore fast and frugal heuristics – simple rules in the mind's adaptive toolbox for making decisions with realistic mental resources. These heuristics can enable both living organisms and artificial systems to make smart choices quickly and with a minimum of information by exploiting the way that information is structured in particular environments. In this précis, we show how simple building blocks that control information search, stop search, and make decisions can be put together to form classes of heuristics, including: ignorance-based and one-reason decision making for choice, elimination models for categorization, and satisficing heuristics for sequential search. These simple heuristics perform comparably to more complex algorithms, particularly when generalizing to new data – that is, simplicity leads to robustness. We present evidence regarding when people use simple heuristics and describe the challenges to be addressed by this research program. Key Words: adaptive toolbox; bounded rationality; decision making; elimination models; environment structure; heuristics; ignorance-based reasoning; limited information search; robustness; satisficing; simplicity. (shrink)

This article discusses the philosophical relevance of recent computational work on inductive inference being conducted in the rapidly growing branch of artificial intelligence called machine learning.

In recent years there has been a great deal of discussion about the prospects of developing a “naturalized epistemology,” though different authors tend to interpret this label in quite different ways.1 One goal of this paper is to sketch three projects that might lay claim to the “naturalized epistemology” label, and to argue that they are not all equally attractive. Indeed, I’ll maintain that the first of the three – the one I’ll attribute to Quine – is simply incoherent. There (...) is no way we could get what we want from an epistemological theory by pursuing the project Quine proposes. The second project on my list is a naturalized version of reliabilism. This project is not fatally flawed in the way that Quine’s is. However, it’s my contention that the sort of theory this project would yield is much less interesting than might at first be thought. (shrink)

In recent years there has been a great deal of discussion about the prospects of developing a ‘naturalized epistemology’, though different authors tend to interpret this label in quite different ways. One goal of this paper is to sketch three projects that might lay claim to the ‘naturalized epistemology’ label, and to argue that they are not all equally attractive. Indeed, I'll maintain that the first of the three—the one I'll attribute to Quine—is simply incoherent. There is no way we (...) could get what we want from an epistemological theory by pursuing the project Quine proposes. The second project on my list is a naturalized version of reliabilism. This project is not fatally flawed in the way that Quine's is. However, it's my contention that the sort of theory this project would yield is much less interesting than might at first be thought. (shrink)

Human and machine discovery are gradual problem-solving processes of searching large problem spaces for incompletely defined goal objects. Research on problem solving has usually focused on search of an instance space (empirical exploration) and a hypothesis space (generation of theories). In scientific discovery, search must often extend to other spaces as well: spaces of possible problems, of new or improved scientific instruments, of new problem representations, of new concepts, and others. This paper focuses especially on the processes for finding new (...) problem representations and new concepts, which are relatively new domains for research on discovery.Scientific discovery has usually been studied as an activity of individual investigators, but these individuals are positioned in a larger social structure of science, being linked by the blackboard of open publication (as well as by direct collaboration). Even while an investigator is working alone, the process is strongly influenced by knowledge and skills stored in memory as a result of previous social interaction. In this sense, all research on discovery, including the investigations on individual processes discussed in this paper, is social psychology, or even sociology. (shrink)

This paper argues that all discoveries, if they can be viewed as autonomous learning from the environment, share a common process. This is the process of model abstraction involving four steps: act, predict, surprise, and refine, all built on top of the discoverer's innate actions, percepts, and mental constructors. The evidence for this process is based on observations on various discoveries, ranging from children playing to animal discoveries of tools, from human problem solving to scientific discovery. Details of this process (...) can be studied with computer simulations of discovery in simulated environments. (shrink)

The paper consists of a reflexive exercise in which Herbert Simon's views concerning science are applied to his own research. It argues that what connected his ventures into so many different disciplinary domains was a search for complex, hierarchical systems. In the process, the paper establishes a close connection between Simon's insights and his focus on simulation. Instead of simulating Simon on a computer, though, it simulates Simon on paper. This exercise is then contrasted with Simon's own attempts to simulate (...) science. This comparison leads to a reflexive evaluation of Simon's simulations. (shrink)

In this paper the basic aim of the so‐called ‘strong programme’ in the sociology of knowledge is examined. The ‘strong programme’ is considered (and rightly so) as an extreme version of the anti‐realist view of science. While the problem of scientific realism has normally been dealt with from the point of view of the ‘context of justification’ of theories, the paper focuses on the issues raised by law‐discovery. In this context Herbert Simon's views about the existence of a ‘logic of (...) scientific discovery’ are discussed and criticized. The main thesis of the paper is that if the structure of both discovery and prediction is properly understood, then the basic anti‐realist claims become untenable. A fortiori, the ‘strong programme’ appears to be unable to explain some basic features of the structure of science. (shrink)

This paper describes an abductive process model of anomalous data integration. The model makes use of the entrenchment of the current explanation and the probability of alternative explanations. It is hypothesised that increasing confirmation of the anom-aly itself increases the probability of alternative explanations. In an experimental study we found that both the entrenchment of an existing explanation and confirmation of the anomaly clearly influence how people resolve anomalous data. These results are in agreement with the predic-tions of the model.

The way scientific discovery has been conceptualized has changed drastically in the last few decades: its relation to logic, inference, methods, and evolution has been deeply reloaded. The ‘philosophical matrix’ moulded by logical empiricism and analytical tradition has been challenged by the ‘friends of discovery’, who opened up the way to a rational investigation of discovery. This has produced not only new theories of discovery, but also new ways of practicing it in a rational and more systematic way. Ampliative rules, (...) methods, heuristic procedures and even a logic of discovery have been investigated, extracted, reconstructed and refined. The outcome is a ‘scientific discovery revolution’: not only a new way of looking at discovery, but also a construction of tools that can guide us to discover something new. This is a very important contribution of philosophy of science to science, as it puts the former in a position not only to interpret what scientists do, but also to provide and improve tools that they can employ in their activity. (shrink)

A novel experimental paradigm that measured theory change and confidence in participants' theories was used in three experiments to test the effects of anomalous evidence. Experiment 1 varied the amount of anomalous evidence to see if “dose size” made incremental changes in confidence toward theory change. Experiment 2 varied whether anomalous evidence was convergent or replicating. Experiment 3 varied whether participants were provided with an alternative theory that explained the anomalous evidence. All experiments showed that participants' confidence changes were commensurate (...) with the amount of anomalous evidence presented, and that larger decreases in confidence predicted theory changes. Convergent evidence and the presentation of an alternative theory led to larger confidence change. Convergent evidence also caused more theory changes. Even when people do not change theories, factors pertinent to the evidence and alternative theories decrease their confidence in their current theory and move them incrementally closer to theory change. (shrink)

Cognitive studies of science and technology have had a long history of largely independent research projects that have appeared in multiple outlets, but rarely together. The emergence of a new International Society for Psychology of Science and Technology suggests that this is a good time to put some of the latest work in this area into topiCS in a way that will both acquaint readers with the cutting edge in this domain and also give them a hint of its history. (...) One core theme includes how scientists, inventors, and engineers represent and solve problems; another, related theme is the extent to which they distribute and share cognition. Methodologies include fine‐grained studies of historical records, protocols of working scientists, observations and comparisons of engineering science laboratories, and computational simulations designed both to serve as research tools and also to improve scientific problem‐solving. The series of articles will conclude with the Associate Editor’s suggestions for future research. (shrink)

The author describes his efforts to become a participant observer while he was a Program Director at the NSF. He describes his plans for keeping track of his reflections and his goals before he arrived at NSF, then includes sections from his reflective diary and comments after he had completed his two-year rotation. The influx of rotators means the NSF has to be an adaptive, learning organization but there are bureaucratic obstacles in the way.

Narrative accounts misrepresent discovery by reconstructing worlds ordered by success rather than the world as explored. Such worlds rarely contain the personal knowledge that informed actual exploration and experiment. This article describes an attempt to recover situated learning in a material environment, tracing the discovery of the first electromagnetic motor by Michael Faraday in September 1821 to show how he modeled new experience and invented procedures to communicate that novelty. The author introduces a notation to map experiment as an active (...) process in a real-world environment and to display the human agency written out of most narratives. Comparing maps of accounts shows how knowledge-construction depends on narrative reconstruction. It is argued that invention processes can be interpreted in the same way as discovery, and a study is proposed to compare packaging learned skills into demonstration devices with the innovative strategies of inventors such as Edison. If situational knowledge is as important as is claimed, computationalists need to join science studies scholars in coming to grips with nonverbal and procedural aspects of discovery and invention. (shrink)

The ArgumentScientific tools—measurement and calculation instruments, techniques of inference—straddle the line between the context of discovery and the context of justification. In discovery, new scientific tools suggest new theoretical metaphors and concepts; and in justification, these tool-derived theoretical metaphors and concepts are morelikely to be accepted by the scientific community if the tools are already entrenched in scientific practice.Techniques of statistical inference and hypothesis testing entered American psychology first as tools in the 1940s and 1950s and then as cognitive theories (...) in the 1960s and 1970s. Not only did psychologists resists statistical metaphors of mind prior to the institutionalization of inference techniques in their own practice; the cognitive theories they ultimately developed about “the mind as intuitive statistician” still bear the telltale marks of the practical laboratory context in which the tool was used. (shrink)

Does recent work in the cognitive sciences have any implications for theories or methods employed within the philosophy of science itself? The answer to this question depends first on one’s conception of the philosophy of science and then on the nature of work being done in the various different fields comprising the cognitive sciences. For example, one might think of the philosophy of science as being an autonomous discipline that is both logically and epistemologically prior to any empirical inquiry. If (...) the cognitive sciences are empirical sciences, then research in the cognitive sciences could not have any significant implications for the philosophy of science. And that would be the end of the matter. Logical Empiricism is now typically understood as having exemplified this point of view.More specifically, Logical Empiricism took it for granted (i) that scientific knowledge should be understood as ideally having the structure of a formal logical calculus, and (ii) that the empirical warrant for scientific claims is given by directly observed data together with formal rules which determine the weight of the evidence for or against the particular claims in question. (shrink)

In the last years there has been a great improvement in the development of computational methods for combinatorial chemistry applied to drug discovery. This approach to drug discovery is sometimes called a “rational way” to manage a well known phenomenon in chemistry: serendipity discoveries. Traditionally, serendipity discoveries are understood as accidental findings made when the discoverer is in quest for something else. This ‘traditional’ pattern of serendipity appears to be a good characterization of discoveries where “luck” plays a key role. (...) In this sense, some initial failures in combinatorial chemistry are frequently attributed to a naïf appropriation of a “serendipity model” for discovery (a “serendipity mistake”). In this paper we try to evaluate this statement by criticizing its foundations. It will be suggested that the notion of serendipity has different aspects and that the criticism to the first attempts could be understood as a “serendipity mistake.” We will suggest that “serendipity” strategies, a kind of blind search, can be seen sometimes as a “genuine part” of scientific practice. A discussion will ensue about how this characterization can give us a better understanding of some aspects of serendipity discoveries. (shrink)

Ever since 1956 when details of the Logic Theorist were published by Newell and Simon, a large literature has accumulated on computational models and theories of the creative process, especially in science, invention and design. But what exactly do these computational models/theories tell us about the way that humans have actually conducted acts of creation in the past? What light has computation shed on our understanding of the creative process? Addressing these questions, we put forth three propositions: (I) Computational models (...) of the creative process are fundamentally flawed as theories of human creativity. Rather, the universal power of computational models lies elsewhere: (II) Computational models of particular acts of creation can serve as effective experiments to test universal hypotheses about creative processes and mechanisms; and (III) Computation-based architectures of the creative mind provide metaphorical frameworks that, like all good metaphors, can serve as rich sources of insight into aspects of the creative process. In this paper, we provide evidence for these three propositions by drawing upon some particular episodes in the cognitive history of science, technology, and art. (shrink)

In the twentieth century, no person epitomized more dramatically the “Renaissance mind” than Herbert A. Simon (1916–2001). In aworking life spanning over 60 years, Simon made seminal contributions to administrative theory, axiomatic foundations of physics, economics, sociology, econometrics, cognitive psychology, logic of scientific discovery, and artificial intelligence. Simon's life of the mind, thus, affords nothing less than a “laboratory” in which to observe and examine at close quarters the phenomenon ofmultidisciplinary creativity. In this paper, we attempt to shed some light (...) on the nature of Simon's creativity and the nature of his particular Renaissance mind. In particular, we have attempted here to articulate thecognitive styleunderlying Simon's multidisciplinary creativity. (shrink)