THREE ABDUCTIVE SOLUTIONS TO THE MENO PARADOX WITH INSTINCT, INFERENCE, AND DISTRIBUTED COGNITION. SAMI PAAVOLA and KAI HAKKARAINEN

Transcription

1 To appear in Studies in Philosophy and Education THREE ABDUCTIVE SOLUTIONS TO THE MENO PARADOX WITH INSTINCT, INFERENCE, AND DISTRIBUTED COGNITION SAMI PAAVOLA and KAI HAKKARAINEN University of Helsinki, Finland ABSTRACT This article analyzes three approaches to resolving the classical Meno paradox, or its variant, the learning paradox, emphasizing Charles S. Peirce s notion of abduction. Abduction provides a way of dissecting those processes where something new, or conceptually more complex than before, is discovered or learned. In its basic form, abduction is a weak form of inference, i.e., it gives only tentative suggestions for further investigation. But it is not too weak if various sources of clues and restrictions on the abductive search are taken into account. We present three, complementary versions of abduction: 1) as a sort of guessing instinct or expert-like intuition, where unconscious clues are important; 2) as a form of inference, where a strategic point of view is essential; and 3) as a part of distributed cognition and mediated activity, where the interaction with the material, social, and cultural environment is emphasized. Our starting point is Peirce s own notion of abduction, but we broaden the perspective, especially to the direction of distributed cognition. Key Words: Meno paradox, learning paradox, abduction, Peirce, guessing, distributed cognition

2 2 INTRODUCTION It has been maintained that Charles S. Peirce s abduction provides a way of answering the classical Meno paradox (Prawat, 1999; Magnani, 2001, pp. 1-14; also Nesher, 2001), one which brings out basic issues related to inquiry and discovery. There are several ways to represent the paradox, but the central idea is to problematize how inquiry in general is possible. In inquiry Either you know what you are searching for or you do not. If you do know, you already have it, whence inquiry is pointless. And if you do not know, you would not recognize it even if you stumbled on it accidentally; hence, again, inquiry is impossible, pointless. (Nickles, 1981, p. 89) The same problem can be applied to learning. Carl Bereiter (1985) has described a variant of the Meno paradox, a so-called learning paradox, succinctly as follows: To put it most simply, the paradox is that if one tries to account for learning by means of mental actions carried out by the learner, then it is necessary to attribute to the learner a prior cognitive structure that is as advanced or complex as the one to be acquired. (ibid., p. 202) This paradox especially concerns discovering or learning of something new or conceptually more complex than before; how are we to understand that people can learn or discover something that goes beyond what they already know, and something that is more complex than they have known? A related, more practical problem is, What is a rational way of searching for something if we do not know what it is? Do we have any means for understanding something that is conceptually more complex than our existing knowledge? Or to put it still in another way, Is there any logic of discovery, of understanding something new, be it in learning or inquiry in general?

3 3 There are various classical ways suggesting how we humans actually solve this paradox, invoking basic epistemological approaches to learning and inquiry, that is, rationalism, and empiricism (Piattelli-Palmarini, 1980), or evolutionary analogies (Cziko, 1995). Roughly, rationalism says that we human beings already have all the necessary conceptual structures with our reason, and to discover or learn something is to recollect or see something that we already have possessed in some form all the time. According to empiricism, the new emerges when observations and experiences connect together often enough, or lend support to an inductive hypothesis. Evolutionary analogies are based on the idea that the new emerges by random variations combined with a strict selection process. Abduction seems to provide yet another alternative for framing the Meno problem because it goes right to the heart of the problem; that is, it is meant to conceptualize a logic of discovery (Hanson, 1958). But at the same time it is also subject to those problems that are brought out in the Meno/learning paradox: It seems to require, at the beginning of the process, such structures as would properly be an outcome of abduction (more on these problems, later). In line with these issues, creation of novelty has largely remained an unexamined area in the philosophy of science because it has often been maintained that the context of discovery is something that cannot be tackled with conceptual means. In this paper we delineate three interpretations of abduction and how they shed light on the Meno puzzle. A central key is that abductive inference is a weak form of inference. It is a way of searching for plausible or possible candidate ideas. It is not, however, hopelessly weak if there are some constraints and helps for the search. First, we delineate briefly how abduction that relates to a guessing instinct can help solve

4 4 the Meno problem. Then we analyze those ways abduction as a form of inference can enlighten the area of discovery. And thirdly, we consider that abductive search for new ideas might be seen as connected with theories concerning distributed cognition. When understood as a part of actual problem-solving processes of human inquirers, the abductive search for ideas in inquiry and learning is a part of an on-going activity where cultural, collaborative, and material means constrain and instigate the search for new ideas. Our focus in this paper is on general epistemological issues; but in the conclusions we will briefly comment on the meaning of abduction more specifically for learning. ABDUCTIVE INSTINCT (OR INTUITION) AS A GUIDE Peirce himself, especially in his later writings, maintained that abduction is a weak mode of inference, that is, it provides only plausible or candidate hypotheses which should be tested with deductive and inductive means (CP , 1908; CP , 1901). According to Peirce, abduction is a form of inference which is near to guessing, or even the same as guessing (Peirce, 1985, pp , 1900; CP 7.219, 1901). But if abduction is so weak, is it more like a random search than a reasoned process? But according to Peirce, discovery cannot be a random process, a point which is shown by the history of science, and by the rapid development of human culture (CP 7.220, 1901; CP 5.591, 1903). Peirce reasoned that there simply would not have been enough time in history for human beings to find successful ideas by pure luck; the boundaries of a solution set are too wide. A physicist, when meeting some new phenomenon should in principle take into account the conjunctions of the planets, or

5 5 the doings of the dowager empress of China (CP 5.172, 1903; also CP 5.591, 1903). Peirce maintained that the chance explanation cannot be rescued even if we presume that scientists have by degrees learned to restrict their search for certain kinds of explanations; there would still not be enough time. Peirce s own explanation, in his later writings, was to maintain that abductive reasoning has its basis on (or is even the same as) some sort of an instinct for finding good hypotheses (CP 7.220, 1901; 6.475, 1908). Peirce gave several formulations for this instinct, which leave room for various interpretations, ranging from metaphysical to naturalistic grounds for this supposed instinctual ability (Shanahan, 1986). An idealistically sounding, and a bit mysterious argument is that when man s mind is developed under the same laws which govern the universe, it is to be expected that man s mind also has a natural tendency, or instinctive insight for guessing those laws themselves (CP 5.604, 1903; cf. CP 6.10, 1891). Peirce also offered various naturalistic arguments for this kind of an instinct. The basic argument was that since animals have instincts which are crucial for their survival why not to assume that we humans have a tendency to find true theories; because this is something that seems to be crucial for our survival (CP 5.591, 1903; also CP 6.531, 1901; CP , c. 1907). Peirce then had several, related grounds for proposing an abductive guessing instinct (Paavola, 2005). One basic proposal is described by Peirce in his famous article Guessing, where he describes his own experience as a detective (Peirce, 1929; partly published in CP , c. 1907; see also Eco & Sebeok, 1983). In this colourful narrative, Peirce tells how he managed to catch a crook and get back his stolen things by successfully using a guessing instinct. According to Peirce, this instinct has its basis on the fact that we often derive from observation strong intimations of truth, without

6 6 being able to specify what were the circumstances we had observed which conveyed those intimations (Peirce 1929, p. 282 [CP 7.46, c. 1907]). The idea is that human beings are able to abduce good and true hypotheses by using clues which they do not recognize consciously. Abduction is closely related to perceptual judgments (see CP , 1903; Hanson, 1958). This type of interpretation makes the abductive instinct appear to be what in cognitive psychology, nowadays, would be called intuition (Hogarth, 2001; Dreyfus & Dreyfus, 1986). Guessing instinct means that one finds a solution or sees that it is right even before any clear evidence for it is recognized. Peirce himself opposed intuition as a starting point for the theory of knowledge if intuition is understood, in a Cartesian manner, as an unanalyzable starting point that is supposed to be independent of any previous knowledge. But at the same time, Peircean guessing instinct has its basis on some non-conscious sign-processes (CP , 1868). From the point of view of modern cognitive science, an intelligent activity of human beings leans on a tremendous knowledge base which can include knowledge of a larger number of situations (50, , 000) than there are words in the human vocabulary (Dreyfus & Dreyfus, 1986). With the development of expertise and experience, human beings develop their skills and abilities to make minute and highly automated distinctions and inferences. Domain experts have an extremely rich knowledge structure that involves knowledge of a large number of exceptional cases, allowing an expert to find a solution to an atypical or difficult case (Feltovich et al., 1997). People have a tendency, in investigation, to strive for a coherent conception where those pieces of information that do not fit to this conception (or to other knowledge) will emerge clearly to the consciousness (Harman, 1986; Thagard, 2000). When human beings develop expertise

7 7 on some specific field their tacit knowledge (Polanyi, 1966), hard to express explicitly but which is crucial for expertise, develops at the same time. This type of knowledge has clear connections to abductive instinct in unconsciously taking account of signs and clues. It has been said that creative experts have a capacity for distinguishing promising ideas from unpromising ones (Bereiter & Scardamalia, 1993). Promisingness means that creative experts are, because of their experience, good at noticing new ideas and seeing very quickly any new directions that may well be important for that field. And creative experts are also constantly trying to find these new ideas and directions. A creative person who has just caught the scent of a promising idea acts in many ways like a bird dog that has just caught the scent of a pheasant. Frozen in motion, ears pricked up, nose turned toward the wind. (Bereiter & Scardamalia, 1993, p. 139) The evaluation of promisingness is not infallible (experts can make big mistakes, just because they rely so much on their existing expertise) but still, it helps one to search for new, promising ideas. Developing this kind of tacit knowledge may be considered as the cultivation of the guessing instinct. Both rest upon the ability to make minute distinctions and know of various alternatives that have been proposed. This kind of an expertise develops bit by bit when the person indwells (see Polanyi, 1966) his research area long enough, adopting various perspectives. So this capability is something that develops with experience and expertise. In fact, there are parallels between Peirce s account of abduction and Polanyi s account of tacit knowledge, especially in relation to the logic of discovery (Mullins, 2002; see also Magnani, 2001, pp. 7-11). We are not maintaining that this kind of a dynamic view of expertise is what Peirce himself had in mind with the guessing instinct. It seems that Peirce was more interested

8 8 in this instinct itself than its development as such. Rather we are maintaining that this is one, rational--and promising--way to interpret Peirce s guessing instinct through modern cognitive science. And this interpretation is at least not explicitly contrary to Peirce s own views because as a pragmatist and evolutionary metaphysician, Peirce emphasized that habits and instincts are subject to change and development (see CP 2.170, c. 1902). ABDUCTIVE SEARCH WITH INFERENTIAL MEANS Abductive instinct/intuition is supposed to be at least a partial answer to the Meno puzzle. So although we do not know the solution to our problem, we can have some clues and impressions that help us to find it, especially if we are familiar with the research area in question. But it seems that this cannot be a complete solution to the Meno paradox. How is it that we understand those new ideas if they are conceptually more complex than before? Can instinct suggest something that is totally new to us? We think that there is room and need for abductive inference in addition to an abductive guessing instinct. In our interpretation, abductive inference has its basis on consciously held premises and operations whereas in abductive instinct, clues and processes that suggest new ideas can be otherwise analogous to inference, but partially unconscious (Paavola, 2005; cf. Thagard, 2000, pp. 2-3). Abductive inference seems, however, also to have problems with the Meno/learning paradox. A basic formulation of abduction is (CP 5.189, 1903): The surprising fact, C, is observed; But if A were true, C would be a matter of course, Hence, there is reason to suspect that A is true.

9 9 But can this be a model for discovery? It appears that the hypothesis searched for ( A, above) is already supposed to be known--or somehow recognized--in the premises. Are we not in the Meno situation again: We are already supposed to know something that we should be just seeking after? Another problem is that abduction is so weak, or wild a form of inference; near to guessing. It seems to allow inferences to almost anything. With a little bit of imagination we can make up all kinds of explanations that would (if true) make the surprising phenomenon a matter of course but this possibility is not a good reason to believe in those fanciful explanations (Achinstein, 1970). Is the above analysis, in the end, the same as saying that we do not have any method for discovery, except maybe good luck, or random trials? We think that abductive model of discovery can be saved, but this means that methodological, or with Peirce s terms Methodeutical, aspects of abduction should be taken into account, not just aspects concerning validity -- Critic in Peirce s terms (see Paavola, 2004a; 2004b). It is important to see abduction also from the point of view of strategical rules. One typical strategy for abductive detectives is to search for anomalous phenomena, i.e., phenomena that are somehow disturbing in the light of what one should have anticipated in relation to one s previous knowledge, but which still seem to be such that they should and could have an explanation. Often it is better if these aberrations are not too big in relation to anticipations because if the anomaly is very big or the phenomenon extremely strange, it does not instigate any inquiry, but rather gives the feeling that this is something that cannot be explained at all. These anomalous phenomena or aberrant details are then used as clues for hypotheses which should explain the phenomena as a matter of course.

10 10 Abduction can conceptualize a logic for discovery if the point is that abductive strategies must guide the process of discovery. This is, we think, why N. R. Hanson wrote in connection to abduction that [a] theory is a cluster of conclusions in search of a premiss (1958, p. 90). Abductive process, in a way, starts with the phenomena to be explained, and it is the whole process of search that is important, not just the relationship between premises and the conclusion. Even if the hypothesis is already in the premises (in the formula of abduction) it is not usually the case that the inquirer recognizes -- beforehand -- that this hypothesis in the premises is applicable as a solution to his/her problem. Or, to put it differently, the particular abductive inference requires, in a similar fashion to analogical reasoning, that the inquirer makes an abduction, in a concrete sense of catching or kidnapping, of an idea from some other area of life to explain the unexpected phenomenon encountered. For example, a crucial abduction of Ignaz Semmelweis ( ) that helped him to find a cause for childbed fever was to realize that this disease was brought about by similar causes to those of blood poisoning in general (Semmelweis, 1983; also Lipton, 1991). It was generally known in his time that a wound from a dirty scalpel can cause a blood poisoning with certain kinds of symptoms, and very often a death. Many people also knew the symptoms of the childbed fever. But Semmelweis discovery was to realize that the symptoms of these blood poisoning cases and the symptoms of childbed fever were similar, and to suggest that maybe the cause was also similar; that is, impure matter or cadaverous particles that were brought to women s organs (Semmelweis, 1983, p. 88). So the hypothesis was in itself old, but in relationship to Semmelweis puzzle it was new.

11 11 This is in line with Howard Gruber s description of Darwin's famous Malthusian insight (Gruber, 1981, p. 42): his [Darwin s] notebooks show that he had or almost had the same idea a number of times before, during the fifteen months of deliberate effort leading up to the moment in question. So the historic moment was in a sense a re-cognition of what he already knew or almost knew. The strategic perspective is even stronger when it is taken into account that normally inquirers are not explaining only one anomalous phenomenon; rather, there are many facts and pieces of information that must, simultaneously, be taken into consideration when new hypotheses are searched for. So even though the hypotheses as such would be old or known, the difficult part might be to understand and construct an explanation that takes into account these various features of the situation. In other words, in relationship to the Meno puzzle, one answer is that the discovery might not need any new conceptual structures as such, but only the ability to understand how to combine or find existing ideas or conceptions in such a way as to solve one s particular problem in question. Strategies entail that various pieces of information and clues guide, and at the same time constrain the search for new hypotheses. Hypotheses are not searched for one by one to explain some individual pieces of information, but the inquirer takes into account that the hypothesis should explain also some other information concerning the subject matter. The inquirer can then tentatively restrict the search for certain types or kinds of hypotheses, and then search for hypotheses that would be subject to this restriction (and there can be many of these tentative restrictions) (Hanson, 1961; Thagard, 1988, pp ; Niiniluoto, 1999, pp. S440-1; Paavola, 2004b). The heuristic selection of hypotheses is also guided by metaconceptual criteria for evaluating explanations (Harman, 1965;

12 12 Thagard, 1988), such as empirical consistency, analogy (acceptable explanations are analogous to well-established scientific theories), non-adhocness (an explanation does not rely on an ad hoc assumptions), and consilience (a hypotheses explains classes of facts that are independent of one another). These principles do not guarantee that a hypothesis is correct, but help to exclude too wild, arbitrary and unconstrained hypotheses. These restrictions are tentative in the sense that they can guide the inquirer to the wrong direction. Still, it can be maintained that these strategic restrictions and explorations are needed, otherwise the inquirer has too many alternatives to try. Yet it might be asked if this is an adequate answer to the Meno/learning paradox? If hypotheses as such need not be new, how were they originally made or discovered? For example, if it was known during Semmelweis time that impure elements in human organs can cause blood poisoning, how was this conception discovered? Are there some elementary conceptions that we should assume as given? We do not claim to have a definite answer to this question. It might, however, be that there are no first conceptions at all, but rather these more elementary conceptions and their origins have a similar abductive structure as more complicated cases (cf. CP 5.181, 1903). In any case, our idea here is to maintain that new conceptual structures do not require that the elements of which they have been constructed are new, but that the area to which they are applied is new or the combination is new, sometimes in a radical way. For example, the Darwinian theory of natural selection is quite complex conceptual structure, which required a conceptual revolution from Darwin and his contemporaries. And it is the same for young students if they are to learn what this theory means. Still, the elements of this theory are not very complex (species generate offspring; these offspring have similar characteristics to their parents; all species very quickly generate more offspring

13 13 that can survive; descendants that have some favourable characteristics are more likely to survive, and so on). Rather the challenge for conceptual change is that these quite simple elements are put together in a certain way so that the pattern of the elements generates the new theory. So although the hypothetical idea itself is in the premises in the formulation of abduction, it is how these ideas are put together in a more complicated way that generates the new theory (see CP 5.181, 1903). ABDUCTION AS A PART OF DISTRIBUTED COGNITION We now delineate in a preliminary manner some of those ways the ideas related to the distributed cognition give important further means to restrict and instigate abductive search. This proposal is related to what Magnani (2001) calls model-based reasoning, and manipulative abduction in contrast to sentential abduction. The abductive model in the previous section is a sentential model; such models present inquiry through propositions and inferences. It is not, however, a very realistic model for describing real-life inquiry processes as such. New ideas do not arise only by making inferences or only in the human head but by acting in concrete, physical environment as a part of some social and cultural community. So inquiry processes must be embedded to be a part of a more general human interaction with the surrounding environment. Modelbased reasoning refers to the construction and manipulation of various kinds of representations, that is, not just sentential representations, but also, for example, visual representations (Magnani, 2001, p. 45; see Thagard & Shelley, 1997). Manipulative abduction is based on manipulation of external objects and representations; it is thinking through doing and not only about doing (ibid., 53). It concerns epistemic

14 14 processes involved in doing inquiry and deliberately interviewing to learn about phenomena rather than just describing them. In recent decades, various models of cognitive psychology have challenged the idea of individuals and individuals' minds as the centre of human activity (Pea, 1993; Hutchins, 1995). According to distributed cognition, mind is not the only source for intelligent, human activity; rather we have to take into account how this activity is culturally, socially, temporally and physically distributed (Salomon, 1993). If we want to understand the dynamics of human activities it is, in a way, a category mistake to focus just on what is happening in individual s minds. The idea of distributed cognition goes well with the perspectives that emphasize the culturally mediated nature of human activities (see Cole & Engeström, 1993). The basic idea is that human activity is mediated by signs and tools which are sort of external artifacts developed by cultural means (Vygotsky, 1978; Engeström et al., 1999). So the focus is not on individual s minds or on structures of mind; rather the starting point is that the human mind also evolves by using and developing culturally mediated tools, signs, and artifacts (Wartofsky, 1979). The focus on mediation, and mediated nature of human activity has clear affinities to Peter Skagestad s way of interpreting Peirce s philosophy. Skagestad (1993) emphasizes evolutionary epistemology, according to which human evolution has happened basically by developing external artifacts and tools, such as written language, sign systems, books, or computers. In order to understand human evolution, or human activities, we should understand how these external artifacts are developed and how they function rather than what is happening inside our minds. Human knowledge is not so much in our heads but rather it is embodied in various external artifacts, such as

15 15 books. The emergence of external representations that allowed overcoming the limitations of human working memory had a crucial significance in human cognitive evolution (Donald, 1991). There are clear affinities between this kind of an evolutionary epistemology and Peirce s remarks concerning theory of knowledge (Skagestad, 1993). Peirce emphasized that human thought happens, not so much inside our heads or minds, but by using signs that are like external artifacts or tools: In my opinion it is much more true that the thoughts of a living writer are in any printed copy of his book than they are in his brain. (CP 7.364, c. 1902; cf. CP 5.289, n.1, 1868; CP 2.54, c. 1902) Ransdell (2003) has added that this externality of signs means that signs are interpreted in dialogical (or communicational) processes. A basic feature of signs is that they are reinterpretable (CP 5.138, 1903; Skagestad, 1993). This interpretation can be done by individuals, but in a fundamental way signs are interpreted and developed in social processes and for communal use. This re-interpretability also means that signs can be interpreted in a long term processes, both by individuals, but also beyond individuals; which is the basis for cultural mediation (cf. e.g., CP 5.289, 1868). We think that distributed cognition and mediated activity are especially important issues to take into account in abduction because it is so weak a form of inference. The elements or ingredients for abductive processes come from the interaction with material, social, and cultural environment. Abduction is a way of searching for promising hypotheses, which are fundamentally only possibilities. In this sense, it is a process where various sorts of clues and restrictions and existing ideas can direct the search in important ways. If we consider abductive search in vivo, that is, how it happens in real life settings and not just as a pure form of inference, these various elements that guide the process are especially important.

16 16 Nesher (2001) has highlighted the interaction with the environment as a basis for abductive processes of discovery. We interact with our environment not just via symbols and symbolic representations, but also via indexical and iconical sign processes (see also Prawat, 1999, pp ). According to Peirce, abductive inference shades into perception and instinctual processes without any sharp line (see CP 5.181, 1903). As we interpret this lack of sharp boundary, it means that human beings have a capability to discern fine-grained clues and to use them in processes that are similar to abductive inference. Another side of this issue is that inquirers get clues and hints in this process by being in interaction with the concrete environment. So to get good abductive ideas, it is not enough to be a good abductive reasoner, but to indwell thoroughly to the research area in question, and to get clues and hints, that is, material for hypotheses. Minute observations and good research instruments are crucially important because they are ways of noticing these small clues. We may say that concrete environment provides affordances (McGrenere & Ho, 2000) for hypotheses. That is, the properties and features of things themselves direct how they can be interpreted. Another direction that restricts and instigates the search for new ideas is cultural practices and cultural-historically developed ways of thinking about the subject matter in question. It appears that Peirce himself did not acknowledge these cultural-historical issues in relation to abduction; at least, not explicitly. As was mentioned above, Peirce maintained that it is not a sufficient explanation for explaining scientists ability to find true theories to think that scientists have learned step by step to exclude certain kinds of un-scientific ways of explaining things and to use explanations that we consider scientific (CP 5.172, 1903; CP 5.591, 1903; cf. CP 7.447, 1893). We agree with Peirce

17 17 that this is not as such a sufficient explanation for our abductive success. But still, these kinds of processes are certainly important in our abductive search. In the Peircean framework, conceptions refer--simultaneously--indexically to reality independent of us as well as collaborative interpretations of these conceptions (see e.g., CP , 1878; CP , 1905). Our conceptions are not only in dialogue with fellow inquirers or with the object of inquiry but always in relation to both of these poles. Thomas Kuhn (1970) called collectively shared ways of seeing and understanding the world in some specific research community paradigms. Ludwik Fleck (1935) used the concept thought style or thought collective to highlight how phenomena that are brought into being and interpreted in some specific historical time are molded crucially by the social community in question. According to Fleck, concepts have histories that affect how we observe and interpret phenomena. It is debatable how compelling these reigning paradigms or thought styles are. We are not taking any general stance on this question here, but certainly existing ways of interpreting and doing things direct the search for new ideas. The reason why modern scientists, when facing anomalous phenomena, do not try to explain them with doings of dowager empress of China (see above), is that these kinds of explanations are not considered apposite in modern scientific communities. The existing ways of interpreting phenomena, and various restrictions are often so compelling that instead of having the problem of choosing from large number of various alternative hypotheses which is the problem of underdetermination often emphasized by philosophers of science, working scientists rather have the problem of finding even one attractive explanation for some phenomena (Lipton, 1991, p. 146).

18 18 In processes of discovery, these existing ways of interpreting things are of course often just those that should be broken, but not all of them at once, otherwise there is no ways to avoid the Meno puzzle. All kinds of existing ways of interpreting things direct the search for new ideas. For example, evolutionary theories are nowadays usually ranked high from scientific point of view, and so it is natural that these theories are used and applied for explaining various phenomena in many research areas. Even in revolutionary scientific changes, where the existing ways of understanding things change radically, not everything is changed at once, but existing ways of interpreting things give elements for new ideas. Our point is that culturally prevailing ideas, even when they do not determine new ideas, give important elements and ideas for the abductive search; even in those cases where the dominant thought style leads scientists to search for theories that are not viable in the long run. One further aspect of the mediated nature of human activity is that ideas and hypotheses developed with abductive processes are not just processed within human minds; they are external, or conceptual artifacts that have their existence and influence also outside human minds (cf. Popper, 1972; Wartofsky, 1979; Skagestad, 1993; Bereiter, 2002). One abductive challenge for communities who are collaboratively developing something new is to transform tacit knowledge of individuals (see above) or knowledge embedded in social practices to external form. In that way, individuals insights and expertise is available for a shared use, and subsequent developmental work (Hakkarainen, et al., 2004; see also Magnani, 2001, p ). External artifacts are represented in various forms, in books, in visual representations, in models, for example. These representations have an influence on our ways of thinking, but only if the community members have access to them. On the other hand, in order to have an

19 19 influence, scientists must put their ideas in such an external form that these ideas are understandable and accessible for other people. Scientific abduction is not just a reasoning process within the human mind; rather, it produces external artifacts, such as journal articles or graphs, which can be used by other researchers (as sources of premises) to support their inquiry processes. One quite obvious source for clues for new ideas is social interaction. We develop new ideas by using other people s ideas, or some elements of these ideas as building blocks for our own hypotheses. This is eloquently described in a citation by Rumelhart et al.: One of the great joys of science lies in the moment of shared discovery. One person's halfbaked suggestion resonates in the mind of another and suddenly takes a definite shape. An insightful critique of one way of thinking about a problem leads to another, better understanding. An incomprehensible simulation result suddenly makes sense as two people try to understand it together. (Rumelhart et al., 1986, p. ix ) In an abductive model, new ideas emerge by taking various clues and restrictions into account, and by searching and combining existing ideas in novel ways. It is very natural to think that this happens within social interaction. In philosophy of science, it is often remarked that in the end it is a scientific community that justifies hypotheses and theories proposed. But also the process of discovery is social from the start (see Paavola, 2004a). Even if new ideas are developed apparently by an individual, this process employs means that are culturally developed, and is in relationship to what other people have said or written in that subject area. Especially if the model of abductive search for hypotheses is broadened to include the comparison between various alternative candidates, that is, the model of inference to the best explanation (Lipton, 1991), it is something that usually happens within social interaction.

20 20 An important aspect of the distributed and mediated nature of human activity is the time-scale where inquiry or learning is seen to happen. One common way of describing methodology is to present it as a cycle; for example, first an abductive generation of the hypothesis, then drawing of deductive consequences, and then an inductive test, and if the test result is negative the cycle starts anew (cf. CP , 1901). As a broad description, this is an apt model for methodology, but it easily gives too linear a view of the processes of inquiry. At least in more difficult problems, the process extends to a long time span, and various clues, ideas, tentative solutions and restrictions are available at the same time, both helping and possibly confusing the search. So the inquirer does not develop hypotheses one by one, but rather has ideas and elements for various sorts of solutions at the same time. N. R. Hanson described this phenomenon in terms that the inquirers have always pressed on for new abductive explanations on the basis of previous and existing explanations and interpretations (Hanson, 1958, pp. 2-3, 88). As an example, he mentions that it took thirty-four years for Galileo to present his constant acceleration hypothesis with confidence; the methodological model should also be able to describe that process, not just the end result (p. 72). Darwin is also famous for his slowly evolving processes of inquiry. So instead of sudden moments of great insights, his discoveries had their basis in long and systematic work with certain objects of research; which processes were then accompanied with many small insights (see Gruber, 1981). This long process, where various pieces and elements are put together to their final form, is an essential part of the process of discovery, not just those rare moments when some big ideas are discovered. Abductive processes of discovery have their basis in the play with possibilities; that is, it is a question of trying to find a pattern

21 21 that puts various pieces of information and background suppositions into a plausible form (cf. Hanson, 1958; Thagard, 2000; Paavola, 2004b). CONCLUSION In this paper, we have presented three kinds of abductive answers to the Meno/learning paradox. A basic answer to the Meno puzzle here is that we have to go between the Menoan horns. The paradox starts from a supposition that we either know something or we do not know that. But this assumption does not take into account the possibility that we can know something in some respects, even when we do not know the whole solution. The abductive answer is to maintain that we must have some, or often quite a lot of elements or ingredients for the new idea before we can really have it (cf. MacMillan & Garrison, 1988, pp ). New ideas emerge from the interaction between different kinds of knowledge and sources of knowledge. Instead of traditional empiricist or rationalist (or selectionist) solutions to the Meno paradox, this abductive approach is a kind of interactionist or semeiotic solution, where various kinds of sign processes (symbolic, indexical, iconic) mediate the search for new ideas. Puzzling phenomena and observations, tacit knowledge, clues from the material world and from the research object, existing ideas, cultural knowledge and tools, social interaction--all these provide material and give direction for novel ideas. Peirce s own notion of abduction is often interpreted as an inferential model, or as a model which emphasizes an instinctual way of generating new ideas (Paavola, 2005). We think that real life context requires that abduction be embedded in processes emphasized in Peirce s system more generally, for example in his pragmaticism,

22 22 anticartesianism, and semiotic theory. For Peirce, human activity is essentially mediated by signs and sign processes. This has interesting parallels to L. S. Vygotsky s ideas of mediated nature of human activity by tools and signs (see Vygotsky, 1978; Engeström, 1987, pp ). Peirce s model of mediation does not explicitly emphasize interaction with the material and cultural environment, as is the case with Vygotsky s tradition. But in this paper we have tried in a preliminary fashion to show that there is room for these elements in abductive models of inquiry. We have proposed that even if abductive processes are analyzed from inferential (and also instinctual) point of view, it is essential to take into account how inquirers get clues, various ideas, and restrictions for their abductive search. Here the interaction with material, social, and cultural environment is crucial. In this paper our focus has been on abductive processes in general, not so much on learning as such. But these processes are applicable to issues of learning (see also e.g., Cunningham, 1998; Prawat, 1999), especially if learning is seen, as we do, in many ways analogous to processes of inquiry (See Hakkarainen & Sintonen, 2002; Hakkarainen et al., 2004). In order to overcome the Meno puzzle, or the learning paradox, learners should have an opportunity to engage into similar sustained processes of collective inquiry that characterize scientific research communities. While some sort of an abductive ability may be a natural aspect of human cognition, an engagement in controlled and disciplined processes of abduction requires specific support and guidance. Learning is not just about knowledge acquisition but also participation in certain kinds of cultural and social practices, and about collaboratively creating and

23 23 developing shared artifacts, practices and knowledge objects in long-term processes (Hakkarainen, et al., 2004). 1 REFERENCES Achinstein, P. (1970). Inference to Scientific Laws. In R. H. Stuwer (ed), Minnesota Studies in the Philosophy of Science 5, Bereiter, C. (1985). Towards a Solution of the Learning Paradox. Review of Educational Research, 55(2), Bereiter, C. (2002). Education and mind in the knowledge age. Hillsdale, NJ: Erlbaum. Bereiter, C. & Scardamalia, M. (1993). Surpassing ourselves. An Inquiry into the Nature and Implications of Expertise. Chicago: Open Court. Cole, M. & Engeström, Y. (1993). A cultural-historical approach to distributed cognition. In G. Salomon (ed) Distributed cognitions: Psychological and educational considerations (pp. 1-46). Cambridge: Cambridge University Press. CP (volume.paragraph, year) see: Peirce, C. S. ( ) Cunningham, D. J. (1998). Cognition as Semiosis: The Role of Inference. Theory and Psychology 8, Cziko, G. (1995). Without Miracles. Universal Selection Theory and the Second Darwinian Revolution. Cambridge, MA: A Bradford Book, The MIT Press. 1 We would like to thank Hal White, M.A. for helpful comments on this paper.

24 24 Donald, M. (1981). The origins of the modern mind. Cambridge, MA: Harvard University Press. Dreyfus, H. & Dreyfus, S. (1986). Mind over machine: The power of human intuition and expertise in the era of the computer. Oxford: Basil Blackwell. Eco, U. & Sebeok, T. A. (Eds) (1988). The Sign of Three. Dupin, Holmes, Peirce. Bloomington: Indiana University Press. Engeström, Y. (1987). Learning by expanding. Helsinki: Orienta-Konsultit. Engeström, Y., Miettinen, R. & Punamäki, R.-L. (Eds) (1999). Perspectives on Activity Theory. Cambridge, MA: Cambridge University Press. Feltovich, P. J., Spiro, R. J. & Coulson, R. L. (1997). Issues of expert flexibility in contexts characterized by complexity and change. In P. J. Feltovich, K. M. Ford, & R. R. Hoffman (eds) Expertise in Context (pp ). Menlo Park, CA: AAAI Press. Fleck, L. (1935). Genesis and Development of a Scientific Fact. Chicago: The University of Chicago Press. Gruber, Howard E. (1981). On the relation between 'aha experiences' and the construction of ideas. History of Science 19, Hakkarainen, K., Palonen, T., Paavola, S. & Lehtinen, E. (2004). Communities of networked expertise: Professional and educational perspectives. Amsterdam: Elsevier. Hakkarainen, K. & Sintonen, M. (2002). Interrogative Approach on Inquiry and Computer-Supported Collaborative Learning. Science & Education 11(1), Hanson, N. R. (1958). Patterns of Discovery. Cambridge: University Press.

25 25 Hanson, N. R. (1961). Is there a logic of scientific discovery. In H. Feigl & G. Maxwell (eds), Current Issues in the Philosophy of Science. New York: Holt, Rinehart and Winston, Inc. Harman, G. (1965). Inference to the best explanation. Philosophical Review 74, Harman, G. (1986). Change in View. Cambridge, MA: MIT Press. Hogarth, R. M. (2001). Educating Intuition. Chicago: The University of Chicago Press. Hutchins, E. (1995). Cognition in the wild. Cambridge, MA: MIT. Kuhn, T. S. (1970). The Structure of Scienti c Revolutions. Second Edition, Enlarged. Chicago: University of Chicago Press. Lipton, P. (1991). Inference to the Best Explanation. London: Routledge. Macmillan, C. J. B. & Garrison, J. W. (1988). A Logical Theory of Teaching. Erotetics and Intentionality. Dordrecht: Kluwer Academic Publishers. Magnani, L. (2001). Abduction, Reason, and Science. Processes of Discovery and Explanation. New York: Kluwer Academic / Plenum Publishers. McGrenere, J. & Ho, W. (2000). Affordances: Clarifying and evolving a concept. Proceedings of Graphics Interface 2000, Available: Mullins, P. (2002). Peirce s Abduction and Polanyi s Tacit Knowing. The Journal of Speculative Philosophy, 16(3), Nesher, D. (2001). Peircean Epistemology of Learning and the Function of Abduction as the Logic of Discovery. Transactions of the Charles S. Peirce Society, 37(1), Nickles, T. (1981). What Is a Problem That We May Solve It?. Synthese 47,

26 26 Niiniluoto, I. (1999). Defending abduction. Philosophy of Science 66: S436-S451. Paavola, S. (2004a). Abduction through Grammar, Critic and Methodeutic. Transactions of the Charles S. Peirce Society, 40(2), Paavola, S. (2004b). Abduction as a Logic of Discovery: The Importance of Strategies. Foundations of Science 9(3), Paavola, S. (2005). Peircean abduction: instinct, or inference? Semiotica 153-1/4, Pea, R. D. (1993). Practices of distributed intelligence and designs for education. In G. Salomon (ed), Distributed cognitions: Psychological and educational considerations (pp ).Cambridge: Cambridge University Press. Peirce, C. S. ( ) [CP (volume.paragraph, year)]. Collected Papers of Charles Sanders Peirce, vol 1-6, C. Hartshorne & P. Weiss (eds), vol 7-8, A. W. Burks (ed). Cambridge, Mass.: Harvard University Press. Peirce, C. S. (1985). Historical Perspectives on Peirces Logic of Science. A History of Science, 2 Vols., C. Eisele (ed). Berlin: Mouton Publishers. Peirce, C. S. (1929). Guessing. Hound & Horn 2(3), Piattelli-Palmarini, M. (Ed.) (1980). Language and Learning. The Debate between Jean Piaget and Noam Chomsky. Cambridge, MA: Harvard University Press. Polanyi, M. (1966). The Tacit Dimension. Gloucester, MA: Peter Smith, reprinted Popper, K. (1972). Objective knowledge: An evolutionary approach. Oxford: Oxford University Press.

27 27 Prawat, R. S. (1999). Dewey, Peirce, and the Learning Paradox. American Educational Research Journal, 36(1), Ransdell, J. (2003). The relevance of Peircean semiotic to computational intelligence augmentation. S.E.E.D. Journal (Semiotics, Evolution, Energy, and Development), 3(3), Available: Rumelhart, D. E. & McClelland, J. L. & the PDP Research Group (1986). Parallel Distributed Processing. Exploration in the Microstructure of Cognition. Volume 1: Foundations. Cambridge, MA: MIT Press. Salomon, G. (Ed) (1993). Distributed cognitions: Psychological and educational considerations. Cambridge: Cambridge University Press. Semmelweis, I. (1983). The Etiology, Concept, and Prophylaxis of Childbed Fever. Translated by K. Codell Carter.Wisconsin: The University of Wisconsin Press. Shanahan, Timothy (1986). The First Moment of Scientific Inquiry: C. S. Peirce on the Logic of Abduction, Transactions of the Charles S. Peirce Society 22(4), Skagestad, P. (1993). Thinking with machines: Intelligence Augmentation, Evolutionary Epistemology, and Semiotic. The Journal of Social and Evolutionary Systems, 16(2), Thagard, P. (1988). Computational philosophy of science. Cambridge, MA: MIT Press. Thagard, Paul (2000). Coherence in Thought and Action. Cambridge, Massachusetts: A Bradford Book, MIT Press.

28 28 Thagard, P. & Shelley, C. (1997). Abductive reasoning: logic, visual thinking, and coherence. In M.L. Dalla Chiara, K. Doets, D. Mundici & J. van Bentham (eds), Logic and Scientific Methods (pp ). Dordrecht: Kluwer. Vygotsky, L. S. (1978). Mind in Society. The Development of Higher Psychological Processes. Cambridge, Massachusetts: Harvard University Press. Wartofsky, M. (1979). Models, Representation, and the Scientific Understanding. Boston: Reidel.