Abstract



What Is A Good Prediction?…an intellectual trait widely considered a great asset in science – the commitment to parsimony – can be a substantial liability in real-world forecasting exercises. ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"pLNstJTR","properties":{"formattedCitation":"(Tetlock 2005, 68)","plainCitation":"(Tetlock 2005, 68)"},"citationItems":[{"id":525,"uris":[""],"uri":[""],"itemData":{"id":525,"type":"book","title":"Expert Political Judgement - How Good Is It? How Can We Know?","publisher":"Princeton University Press","publisher-place":"Princeton and Oxford","event-place":"Princeton and Oxford","author":[{"family":"Tetlock","given":"Philip"}],"issued":{"date-parts":[["2005"]]}},"locator":"68"}],"schema":""} (Tetlock 2005, 68)AbstractPrediction has not received treatment as a philosophical topic in its own right. It has been discussed in connection with explanation, realism, and other topics. But the question “What is a good prediction?” has not been the focus of these treatments. This paper aims to address that question directly. I set out some central distinctions and stances on prediction: between testing and forecasting predictive contexts; between temporal and epistemic notions of prediction; between prediction claims and activities; and between predictive goodness and truth. I go on to argue that there is no ready-made theory of prediction in terms of extrapolation, laws of nature, causation, or mechanisms, because there are circumstances in which each produces bad predictions. I review some psychological evidence concerning cognitive styles associated with predictive accuracy; the better style seems to be “fox-like”, as opposed to “hedgehog-like”. Finally I sketch a philosophical theory of good prediction: a good predictive activity secures warrant for its claim by identifying and verifying currently observable consequences of the prediction claim it supports, where “currently observable consequences” need to meet certain specified conditions. This account also elucidates the plagued question of how exactly prediction and explanation are related.Prediction as a Neglected TopicPhilosophers have given remarkably little thought to prediction, and almost none to prediction as a philosophical topic in its own right. The lack is particularly striking in the philosophy of science, where one might expect prediction to be a central topic. Yet it is not. Undergraduate courses cover confirmation, explanation, laws of nature, and causation; but the only questions commonly asked about prediction are whether “novel predictions” have special confirmatory significance, and whether prediction stands in a close logical relation to explanation. The question, “What is a good prediction?” does not range alongside “What is a good explanation?”, “What is causation?” and the rest. There are no canonical theories of prediction – neither descriptive theories of what in it fact is, nor prescriptive theories of what it must be like to be good. No great philosopher has staked out a view of the topic. Carl Hempel studied the logic of confirmation and of explanation ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"QTzICJK7","properties":{"formattedCitation":"(C. G. Hempel 1945a; C. G. Hempel 1945b; C. Hempel and Oppenheim 1948)","plainCitation":"(C. G. Hempel 1945a; C. G. Hempel 1945b; C. Hempel and Oppenheim 1948)"},"citationItems":[{"id":490,"uris":[""],"uri":[""],"itemData":{"id":490,"type":"article-journal","title":"Studies in the Logic of Confirmation (I.)","container-title":"Mind","page":"1-26","volume":"54","author":[{"family":"Hempel","given":"Carl G."}],"issued":{"date-parts":[["1945"]]}}},{"id":521,"uris":[""],"uri":[""],"itemData":{"id":521,"type":"article-journal","title":"Studies in the Logic of Confirmation (II.)","container-title":"Mind","page":"97-121","volume":"54","author":[{"family":"Hempel","given":"Carl G."}],"issued":{"date-parts":[["1945"]]}}},{"id":481,"uris":[""],"uri":[""],"itemData":{"id":481,"type":"article-journal","title":"Studies in the Logic of Explanation","container-title":"Philosophy of Science","page":"135-175","volume":"15","issue":"2","author":[{"family":"Hempel","given":"Carl"},{"family":"Oppenheim","given":"Paul"}],"issued":{"date-parts":[["1948"]]}}}],"schema":""} (C. G. Hempel 1945a; C. G. Hempel 1945b; C. Hempel and Oppenheim 1948), but he did not write “Studies in the Logic of Prediction”. Prediction is discussed in both the aforementioned studies, but again only incidentally, in relation to the central topic.Given this picture of philosophical neglect, it is an easy matter to read much of what philosophers of science have published on prediction in the last hundred years or so. To summarise that scarce literature, philosophers have discussed:whether prediction is a goal of science ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"XBqbkrku","properties":{"formattedCitation":"(e.g. Whitmore 1942; Salmon 1981; Douglas 2009)","plainCitation":"(e.g. Whitmore 1942; Salmon 1981; Douglas 2009)"},"citationItems":[{"id":548,"uris":[""],"uri":[""],"itemData":{"id":548,"type":"article-journal","title":"The Status of Prediction","container-title":"The Journal of Philosophy","page":"5-14","volume":"39","issue":"1","author":[{"family":"Whitmore","given":"Charles E."}],"issued":{"date-parts":[["1942"]]}},"prefix":"e.g."},{"id":536,"uris":[""],"uri":[""],"itemData":{"id":536,"type":"article-journal","title":"Rational Prediction","container-title":"The British Journal for the Philosophy of Science","page":"115-125","volume":"32","issue":"2","author":[{"family":"Salmon","given":"Wesley C."}],"issued":{"date-parts":[["1981"]]}}},{"id":550,"uris":[""],"uri":[""],"itemData":{"id":550,"type":"article-journal","title":"Reintroducing Prediction to Explanation","container-title":"Philosophy of Science","page":"444-463","volume":"76","author":[{"family":"Douglas","given":"Heather"}],"issued":{"date-parts":[["2009"]]}}}],"schema":""} (e.g. Whitmore 1942; Salmon 1981; Douglas 2009);how predictions are related to explanations, especially when the latter are construed as essentially deductive in character ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"b37Xi6aX","properties":{"formattedCitation":"(e.g. C. Hempel and Oppenheim 1948; Scheffler 1957; Rescher 1958; Canfield and Lehrer 1961; Rescher 1963; Kim 1964; Suchting 1967; Hanna 1969; Douglas 2009)","plainCitation":"(e.g. C. Hempel and Oppenheim 1948; Scheffler 1957; Rescher 1958; Canfield and Lehrer 1961; Rescher 1963; Kim 1964; Suchting 1967; Hanna 1969; Douglas 2009)"},"citationItems":[{"id":481,"uris":[""],"uri":[""],"itemData":{"id":481,"type":"article-journal","title":"Studies in the Logic of Explanation","container-title":"Philosophy of Science","page":"135-175","volume":"15","issue":"2","author":[{"family":"Hempel","given":"Carl"},{"family":"Oppenheim","given":"Paul"}],"issued":{"date-parts":[["1948"]]}},"prefix":"e.g."},{"id":533,"uris":[""],"uri":[""],"itemData":{"id":533,"type":"article-journal","title":"Explanation, Prediction, and Abstraction","container-title":"The British Journal for the Philosophy of Science","page":"293-309","volume":"7","issue":"28","author":[{"family":"Scheffler","given":"Israel"}],"issued":{"date-parts":[["1957"]]}}},{"id":549,"uris":[""],"uri":[""],"itemData":{"id":549,"type":"article-journal","title":"On Prediction and Explanation","page":"281-290","volume":"8","issue":"32","author":[{"family":"Rescher","given":"Nicholas"}],"issued":{"date-parts":[["1958"]]}}},{"id":559,"uris":[""],"uri":[""],"itemData":{"id":559,"type":"article-journal","title":"A Note on Prediction and Deduction","container-title":"Philosophy of Science","page":"204-208","volume":"28","issue":"2","author":[{"family":"Canfield","given":"John"},{"family":"Lehrer","given":"Keith"}],"issued":{"date-parts":[["1961"]]}}},{"id":530,"uris":[""],"uri":[""],"itemData":{"id":530,"type":"article-journal","title":"Discrete State Systems, Markov Chains, and Problems in the Theory of Scientific Explanation and Prediction","container-title":"Philosophy of Science","page":"325-345","volume":"30","issue":"4","author":[{"family":"Rescher","given":"Nicholas"}],"issued":{"date-parts":[["1963"]]}}},{"id":541,"uris":[""],"uri":[""],"itemData":{"id":541,"type":"article-journal","title":"Inference, Explanation, and Prediction","container-title":"The Journal of Philosophy","page":"360-368","volume":"61","issue":"12","author":[{"family":"Kim","given":"Jaegwon"}],"issued":{"date-parts":[["1964"]]}}},{"id":544,"uris":[""],"uri":[""],"itemData":{"id":544,"type":"article-journal","title":"Deductive Explanation and Prediction Revisited","container-title":"Philosophy of Science","page":"41-52","volume":"34","issue":"1","author":[{"family":"Suchting","given":"WA"}],"issued":{"date-parts":[["1967"]]}}},{"id":528,"uris":[""],"uri":[""],"itemData":{"id":528,"type":"article-journal","title":"Explanation, Prediction, Description, and Information Theory","container-title":"Synthese","page":"308-334","volume":"20","issue":"3","author":[{"family":"Hanna","given":"Joseph F."}],"issued":{"date-parts":[["1969"]]}}},{"id":550,"uris":[""],"uri":[""],"itemData":{"id":550,"type":"article-journal","title":"Reintroducing Prediction to Explanation","container-title":"Philosophy of Science","page":"444-463","volume":"76","author":[{"family":"Douglas","given":"Heather"}],"issued":{"date-parts":[["2009"]]}}}],"schema":""} (e.g. C. Hempel and Oppenheim 1948; Scheffler 1957; Rescher 1958; Canfield and Lehrer 1961; Rescher 1963; Kim 1964; Suchting 1967; Hanna 1969; Douglas 2009);how prediction is related to confirmation ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"ou0WDFaH","properties":{"formattedCitation":"(e.g. C. G. Hempel 1945a; Matthew 1971; Merrill 1979; Douglas 2009)","plainCitation":"(e.g. C. G. Hempel 1945a; Matthew 1971; Merrill 1979; Douglas 2009)"},"citationItems":[{"id":490,"uris":[""],"uri":[""],"itemData":{"id":490,"type":"article-journal","title":"Studies in the Logic of Confirmation (I.)","container-title":"Mind","page":"1-26","volume":"54","author":[{"family":"Hempel","given":"Carl G."}],"issued":{"date-parts":[["1945"]]}},"prefix":"e.g."},{"id":557,"uris":[""],"uri":[""],"itemData":{"id":557,"type":"article-journal","title":"Prediction and Predication","container-title":"The British Journal for the Philosophy of Science","page":"171-182","volume":"22","issue":"2","author":[{"family":"Matthew","given":"Anthony"}],"issued":{"date-parts":[["1971"]]}}},{"id":542,"uris":[""],"uri":[""],"itemData":{"id":542,"type":"article-journal","title":"Prediction and Confirmation","container-title":"Philosophy of Science","page":"98-117","volume":"46","issue":"1","author":[{"family":"Merrill","given":"G.H."}],"issued":{"date-parts":[["1979"]]}}},{"id":550,"uris":[""],"uri":[""],"itemData":{"id":550,"type":"article-journal","title":"Reintroducing Prediction to Explanation","container-title":"Philosophy of Science","page":"444-463","volume":"76","author":[{"family":"Douglas","given":"Heather"}],"issued":{"date-parts":[["2009"]]}}}],"schema":""} (e.g. C. G. Hempel 1945a; Matthew 1971; Merrill 1979; Douglas 2009);further, whether predictions offer better evidence for or against theories than empirical claims of non-predictive kinds ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"r5F9i9Lt","properties":{"formattedCitation":"(e.g. C. G. Hempel 1945a; Maher 1988; Maher 1993; White 2003; Lipton 2005; Douglas 2009)","plainCitation":"(e.g. C. G. Hempel 1945a; Maher 1988; Maher 1993; White 2003; Lipton 2005; Douglas 2009)"},"citationItems":[{"id":490,"uris":[""],"uri":[""],"itemData":{"id":490,"type":"article-journal","title":"Studies in the Logic of Confirmation (I.)","container-title":"Mind","page":"1-26","volume":"54","author":[{"family":"Hempel","given":"Carl G."}],"issued":{"date-parts":[["1945"]]}},"prefix":"e.g."},{"id":535,"uris":[""],"uri":[""],"itemData":{"id":535,"type":"article-journal","title":"Prediction, Accommodation, and the Logic of Discovery","container-title":"PSA: Proceedings of the Biennial Meeting of the Philosophy of Science Association","page":"273-285","volume":"1","author":[{"family":"Maher","given":"Patrick"}],"issued":{"date-parts":[["1988"]]}}},{"id":558,"uris":[""],"uri":[""],"itemData":{"id":558,"type":"article-journal","title":"Howson and Franklin on Prediction","container-title":"Philosophy of Science","page":"329-340","volume":"60","issue":"2","author":[{"family":"Maher","given":"Patrick"}],"issued":{"date-parts":[["1993"]]}}},{"id":526,"uris":[""],"uri":[""],"itemData":{"id":526,"type":"article-journal","title":"The Epistemic Advantage of Prediction over Accommodation","container-title":"Mind","page":"653-683","volume":"112","issue":"448","author":[{"family":"White","given":"Roger"}],"issued":{"date-parts":[["2003"]]}}},{"id":454,"uris":[""],"uri":[""],"itemData":{"id":454,"type":"article-journal","title":"Testing Hypotheses: Prediction and Prejudice","container-title":"Science","page":"219-221","volume":"307","issue":"5707","author":[{"family":"Lipton","given":"Peter"}],"issued":{"date-parts":[["2005"]]}}},{"id":550,"uris":[""],"uri":[""],"itemData":{"id":550,"type":"article-journal","title":"Reintroducing Prediction to Explanation","container-title":"Philosophy of Science","page":"444-463","volume":"76","author":[{"family":"Douglas","given":"Heather"}],"issued":{"date-parts":[["2009"]]}}}],"schema":""} (e.g. C. G. Hempel 1945a; Maher 1988; Maher 1993; White 2003; Lipton 2005; Douglas 2009);whether “novel predictions” offer evidence for scientific realism, or can be accommodated by various anti-realisms ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"Mprgawks","properties":{"formattedCitation":"(e.g. Van Fraassen 1980; Laudan 1981; Lawson 1985; Lipton 2004; Lipton 2005)","plainCitation":"(e.g. Van Fraassen 1980; Laudan 1981; Lawson 1985; Lipton 2004; Lipton 2005)"},"citationItems":[{"id":361,"uris":[""],"uri":[""],"itemData":{"id":361,"type":"book","title":"The Scientific Image","publisher":"Oxford University Press","publisher-place":"Oxford","event-place":"Oxford","author":[{"family":"Van Fraassen","given":"Bas"}],"issued":{"date-parts":[["1980"]]}},"prefix":"e.g."},{"id":487,"uris":[""],"uri":[""],"itemData":{"id":487,"type":"article-journal","title":"A Confutation of Convergent Realism","container-title":"Philosophy of Science","page":"19-49","volume":"48","issue":"1","author":[{"family":"Laudan","given":"Larry"}],"issued":{"date-parts":[["1981"]]}}},{"id":555,"uris":[""],"uri":[""],"itemData":{"id":555,"type":"article-journal","title":"The Context of Prediction (and the Paradox of Confirmation)","container-title":"The British Journal for the Philosophy of Science","page":"393-407","volume":"36","issue":"4","author":[{"family":"Lawson","given":"Tony"}],"issued":{"date-parts":[["1985"]]}}},{"id":352,"uris":[""],"uri":[""],"itemData":{"id":352,"type":"book","title":"Inference to the Best Explanation","publisher":"Routledge","publisher-place":"London and New York","edition":"Second edition","event-place":"London and New York","author":[{"family":"Lipton","given":"Peter"}],"issued":{"date-parts":[["2004"]]}}},{"id":454,"uris":[""],"uri":[""],"itemData":{"id":454,"type":"article-journal","title":"Testing Hypotheses: Prediction and Prejudice","container-title":"Science","page":"219-221","volume":"307","issue":"5707","author":[{"family":"Lipton","given":"Peter"}],"issued":{"date-parts":[["2005"]]}}}],"schema":""} (e.g. Van Fraassen 1980; Laudan 1981; Lawson 1985; Lipton 2004; Lipton 2005).Philosophers have not seriously asked how predictions are generated, or how we distinguish, or ought to distinguish, good from bad predictions. Sometimes it seems that they have assumed that predictions are a deductive fall-out of theorising, and are thus generated more or less automatically. It remains common to see it assumed that universal generalisations employing empirical predicates – such as “All ravens are black” – have predictive import, such as “The raven I see at the raven park tomorrow will be black.” First Pierre Duhem and later W.V.O. Quine clearly showed that this assumption is false ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"IDDyWr3n","properties":{"formattedCitation":"(Duhem 1914 Pt II Ch IV; Quine 1953)","plainCitation":"(Duhem 1914 Pt II Ch IV; Quine 1953)"},"citationItems":[{"id":430,"uris":[""],"uri":[""],"itemData":{"id":430,"type":"book","title":"The Aim and Structure of Physical Theory","publisher":"Atheneum","publisher-place":"New York","event-place":"New York","language":"Translated from the French by Philip P. Wiener","author":[{"family":"Duhem","given":"Pierre"}],"issued":{"date-parts":[["1914"]],"season":"1974"}},"suffix":"Pt II Ch IV"},{"id":201,"uris":[""],"uri":[""],"itemData":{"id":201,"type":"chapter","title":"Two Dogmas of Empiricism","container-title":"From a Logical Point of View","publisher":"Harvard University Press","publisher-place":"Cambridge, Mass.","page":"20―43","event-place":"Cambridge, Mass.","author":[{"family":"Quine","given":"W. V.O"}],"issued":{"date-parts":[["1953"]]}}}],"schema":""} (Duhem 1914 Pt II Ch IV; Quine 1953): merely framing a universal generalisation using empirical predicates does not guarantee predictive import; often, perhaps always, auxiliary hypotheses are needed before there is an entailment.Alternatively, philosophers may be silent on prediction because they see it as a psychological matter, like the psychological process of theory generation, and thus an empirical mystery about which they, as philosophers, need not say anything. This is not a stable attitude. The relation of theory to empirical evidence and to superempirical virtures is the bread and butter of philosophers of science. We do not feel shy about saying what makes a good scientific theory, notwithstanding our lack of insight into the spark of genius that produced it. Likewise, we should not feel shy about saying what makes a good prediction, notwithstanding our lack of psychological insight into the psychological processes of predicting. And anyway, these processes are not as mysterious as all that: we will take a cursory glance at some psychological evidence in Section REF _Ref349132476 \r \h IV PAGEREF _Ref349132476 \h below.There are thus no good excuses for ignoring the philosophical sense of the question, “What is a good prediction?” Three points of preamble are in order before proceeding to an answer. First, I am not denying that prediction is discussed in many disciplines outside philosophy: economics, epidemiology, climate science, and so forth. None of these disciplines seek to arrive at necessary and sufficient conditions for a prediction to be good, however. Rather they seek practical, usually domain-specific methods for making and improving predictions. Just as paint manufacturers do not struggle with the question “What is colour?”, sciences engaged in forecasting do not struggle with the question “What is a good prediction?” in the general sense that it is intended here. There is no easy answer outside philosophy, just as there is no easy answers can be had to “What is explanation?” or “What is causation?” by pointing out that many scientific disciplines make use of these notions.Second, there are treatments of prediction by philosophers that are not philosophical in the sense I am seeking. Chief among them is the work of Peter Spirtes, Clark Glymour and Richard Scheines. In the context of causation, they distinguish philosophical from mathematical approaches. Philosophical approaches seek necessary and sufficient conditions; mathematical approaches “seek to provide axioms that use the notion of causation without defining it, and to investigate the necessary consequences of those assumptions” ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"c4x1Pcm7","properties":{"formattedCitation":"(Spirtes, Glymour, and Scheines 2000, 3)","plainCitation":"(Spirtes, Glymour, and Scheines 2000, 3)"},"citationItems":[{"id":168,"uris":[""],"uri":[""],"itemData":{"id":168,"type":"book","title":"Causation, Prediction, and Search","publisher":"Springer-Verlag","publisher-place":"New York, N.Y.","edition":"2","event-place":"New York, N.Y.","author":[{"family":"Spirtes","given":"P."},{"family":"Glymour","given":"C."},{"family":"Scheines","given":"R."}],"issued":{"date-parts":[["2000"]]}},"locator":"3"}],"schema":""} (Spirtes, Glymour, and Scheines 2000, 3). These authors also discuss prediction at length, but their approach is explicitly not philosophical in the sense of providing necessary and sufficient conditions. Similar remarks apply to philosophically-informed work by Judea Pearl and colleagues ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"TBoHCtVY","properties":{"formattedCitation":"(Halpern and Pearl 2005; Pearl 2009)","plainCitation":"(Halpern and Pearl 2005; Pearl 2009)"},"citationItems":[{"id":394,"uris":[""],"uri":[""],"itemData":{"id":394,"type":"article-journal","title":"Causes and Explanations: A Structural Models Approach. Part 1: Causes","container-title":"British Journal for the Philosophy of Science","page":"843-887","volume":"56","author":[{"family":"Halpern","given":"Joseph Y."},{"family":"Pearl","given":"Judea"}],"issued":{"date-parts":[["2005"]]}}},{"id":488,"uris":[""],"uri":[""],"itemData":{"id":488,"type":"book","title":"Causality: Models, Reasoning and Inference","publisher":"Cambridge University Press","publisher-place":"Cambridge","edition":"2","event-place":"Cambridge","author":[{"family":"Pearl","given":"Judea"}],"issued":{"date-parts":[["2009"]]}}}],"schema":""} (Halpern and Pearl 2005; Pearl 2009). Thus although they can justifiably claim to have treated the question of this paper, they explicitly set aside the project of answering it in the traditional philosophical fashion, which they regard as having track record distinctly inferior to that of the mathematical approach ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"Jf47rD1t","properties":{"formattedCitation":"(Spirtes, Glymour, and Scheines 2000, 3)","plainCitation":"(Spirtes, Glymour, and Scheines 2000, 3)"},"citationItems":[{"id":168,"uris":[""],"uri":[""],"itemData":{"id":168,"type":"book","title":"Causation, Prediction, and Search","publisher":"Springer-Verlag","publisher-place":"New York, N.Y.","edition":"2","event-place":"New York, N.Y.","author":[{"family":"Spirtes","given":"P."},{"family":"Glymour","given":"C."},{"family":"Scheines","given":"R."}],"issued":{"date-parts":[["2000"]]}},"locator":"3"}],"schema":""} (Spirtes, Glymour, and Scheines 2000, 3). More power to them; but I want to have a go at the traditional kind of answer.Third, let me acknowledge those philosophical treatments that do take prediction seriously. Nancy Cartwright’s recent work is directly concerned with prediction (and indeed lit the fuse for the present paper). But she does not seek to offer a general answer to the question of this paper, so much as identify false assumptions between predictive activities in policy contexts ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"ro0ZxiLE","properties":{"formattedCitation":"(e.g. Cartwright 2010; Cartwright 2011)","plainCitation":"(e.g. Cartwright 2010; Cartwright 2011)"},"citationItems":[{"id":301,"uris":[""],"uri":[""],"itemData":{"id":301,"type":"paper-conference","title":"Will this policy work for you? Predicting effectiveness better: How philosophy helps","container-title":"PSA Presidential Address","event":"Philosophy of Science Association Biennial Meeting","URL":"","author":[{"family":"Cartwright","given":"Nancy"}],"issued":{"date-parts":[["2010"]]},"accessed":{"date-parts":[["2012",5,1]]}},"prefix":"e.g. "},{"id":83,"uris":[""],"uri":[""],"itemData":{"id":83,"type":"article-journal","title":"Predicting what will happen when we act. What counts for warrant?","container-title":"Preventive Medicine","page":"221-4","volume":"53","issue":"4-5","journalAbbreviation":"Prev. Med.","author":[{"family":"Cartwright","given":"Nancy"}],"issued":{"date-parts":[["2011"]]}}}],"schema":""} (e.g. Cartwright 2010; Cartwright 2011). Another interesting treatment of prediction, coming closer to directly addressing the general question posed in my title, is Heather Douglas’s ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"y5JjC3CI","properties":{"formattedCitation":"(2009)","plainCitation":"(2009)"},"citationItems":[{"id":550,"uris":[""],"uri":[""],"itemData":{"id":550,"type":"article-journal","title":"Reintroducing Prediction to Explanation","container-title":"Philosophy of Science","page":"444-463","volume":"76","author":[{"family":"Douglas","given":"Heather"}],"issued":{"date-parts":[["2009"]]}},"suppress-author":true}],"schema":""} (2009). Before setting off into the wilderness, let us consider whether it answers our question.Douglas starts by making the same point I am making in this section:Despite the fact that most philosophers acknowledge the general importance of prediction for science, the vast majority of the intellectual focus between the two goals rests on explanation. Prediction is rarely a topic in its own right, appearing mainly in discussions of con?rmation, realism, and other topics. It has been this way for over 40 years. ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"dccqpenv","properties":{"formattedCitation":"(Douglas 2009, 445)","plainCitation":"(Douglas 2009, 445)"},"citationItems":[{"id":550,"uris":[""],"uri":[""],"itemData":{"id":550,"type":"article-journal","title":"Reintroducing Prediction to Explanation","container-title":"Philosophy of Science","page":"444-463","volume":"76","author":[{"family":"Douglas","given":"Heather"}],"issued":{"date-parts":[["2009"]]}},"locator":"445"}],"schema":""} (Douglas 2009, 445)Welcome as this sentiment is from my perspective, Douglas goes on to treat prediction as if its main value were derivative on a general understanding of the world:The value of predictions has always been clear: predictions help us check whether our accounts of the world have any veracity and, once proven reliable, help us make better decisions about how to proceed in our lives. ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"d9IFVvAk","properties":{"formattedCitation":"(Douglas 2009, 453)","plainCitation":"(Douglas 2009, 453)"},"citationItems":[{"id":550,"uris":[""],"uri":[""],"itemData":{"id":550,"type":"article-journal","title":"Reintroducing Prediction to Explanation","container-title":"Philosophy of Science","page":"444-463","volume":"76","author":[{"family":"Douglas","given":"Heather"}],"issued":{"date-parts":[["2009"]]}},"locator":"453"}],"schema":""} (Douglas 2009, 453)Notice that she does not say that the value of predictions is that they tell us what is going to happen. It is rather that they provide a vehicle for testing our “accounts of the world”. This is a convoluted, though philosophically common, way to understand the point of prediction, which for the common person is primarily useful for what it can tell us about the object of prediction, namely, the future, and only secondarily for what it can tell us about the tools used to make the prediction.Douglas’s conception of the usefulness of understanding prediction (as opposed to the usefulness of prediction) is likewise derivative:The overemphasis on explanation, and excessive distancing between explanation and prediction, has led us astray. If we reconsider the possibilities of a relationship between the two, particularly beyond the straitjacket of a purely logical relationship, we can get a deeper answer concerning why explanations are important, thus answering more thoroughly Salmon’s query, “Why ask ‘why?’?” [sic] We can also gain insight into which explanations should be taken more seriously in science and which less so. ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"6z7B1BfL","properties":{"formattedCitation":"(Douglas 2009, 445)","plainCitation":"(Douglas 2009, 445)"},"citationItems":[{"id":550,"uris":[""],"uri":[""],"itemData":{"id":550,"type":"article-journal","title":"Reintroducing Prediction to Explanation","container-title":"Philosophy of Science","page":"444-463","volume":"76","author":[{"family":"Douglas","given":"Heather"}],"issued":{"date-parts":[["2009"]]}},"locator":"445"}],"schema":""} (Douglas 2009, 445)Throughout the paper, and especially in the concluding paragraphs, it is clear that a theory of explanation rather than prediction is being presented. Prediction does occupy an important role in the picture, since explanation is presented as, roughly, a tool for generating predictions that may be used to test theories. But the manner in which explanations generate predictions is not the subject of close attention, and the question “What makes a prediction good for testing a theory?” is not broached. The title confirms that explanation is the beneficiary of the discussion: “Reintroducing Prediction to Explanation”.I single Douglas’s (very useful) paper out because it is the closest I have found to an explicit contemporary treatment of prediction. Closest, but still not very close: it does not present a theory of prediction. Douglas does not treat prediction as a philosophical topic in its own right, apart from its links to explanation, notwithstanding her apparent awareness of the possibility and lack of such a treatment.I hope that the remainder of this essay will dispel the idea that there is little philosophical interest in prediction. The next section draws four basic, important distinctions concerning prediction. Section REF _Ref349132450 \r \h \* MERGEFORMAT III considers philosophical resources that might offer ready-made philosophical models of prediction, and finds them all wanting. Section REF _Ref349132476 \r \h \* MERGEFORMAT IV considers empirical evidence concerning the nature of good prediction, drawn Philip Tetlock’s large psychological study on predictive reliability of putative political experts. In Section REF _Ref349132758 \r \h \* MERGEFORMAT V I offer a theory of good prediction.DistinctionsThere are several important distinctions to be drawn concerning different ways one might think about prediction, or alternatively, between different kinds of prediction. Making these distinctions helps us to avoid verbal disputes about whether the word “prediction” is correctly used one way or another. It also helps us see in outline what a theory of good prediction might be like.First, we can distinguish between two contexts in which predictions are made: the context of testing and the context of forecasting. A prediction made in the context of testing is made for the purpose of testing some hypothesis that was used to generate the prediction. It may not be a serious claim about the future (or any other time); it may be entertained entirely hypothetically. A prediction made in the context of forecasting, on the other hand, is a genuine effort to say what is going to happen. Philosophers of science naturally assume the context of testing: Douglas, as we have seen, expresses the practical importance of prediction in terms of what predictions can tell us about our theories. Epidemiologists and economists, on the other hand, naturally assume the context of forecasting, and would express the importance of prediction in terms of what it can tell us, not about the theories that generated the prediction, but about the future.Second, we can distinguish between temporal and epistemic predictions (or temporal and epistemic senses of “prediction”; it makes no difference). A temporal prediction entails a claim about the future. Thus Israel Scheffler takes it as obvious that predictions are bound to temporal locations, and thus cannot be propositions:…note that ‘is a prediction’ is not properly applicable to abstract sentences or propositions, since the same sentence ‘It rains on May 8, 1952’ is or is not a prediction depending on the temporal circumstances of its utterance. ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"X4q5EbQc","properties":{"formattedCitation":"(Scheffler 1957, 295)","plainCitation":"(Scheffler 1957, 295)"},"citationItems":[{"id":533,"uris":[""],"uri":[""],"itemData":{"id":533,"type":"article-journal","title":"Explanation, Prediction, and Abstraction","container-title":"The British Journal for the Philosophy of Science","page":"293-309","volume":"7","issue":"28","author":[{"family":"Scheffler","given":"Israel"}],"issued":{"date-parts":[["1957"]]}},"locator":"295"}],"schema":""} (Scheffler 1957, 295)This line of reasoning relies on the implicit premise that a prediction is about the future of the time of utterance.On the other hand, an epistemic prediction need not entail a claim about the future, but rather entails a claim that (i) falls short of knowledge, and (ii) does so because the truth of the claim has not yet been checked by any more direct means than the means by which it was arrived at in the course of making the prediction.Thus Heather Douglas writes:I… will take prediction to be indexed to the predictor’s epistemological state rather than temporal location. Thus, a claim that we should ?nd a piece of evidence in a particular context is a prediction, even if that context occurred in the past and the evidence is merely preserved and awaiting discovery. Indeed, if someone claims that some event should have taken place, but they do not know that it did, and someone else already does, that should still count as a prediction. ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"wcvlyFB9","properties":{"formattedCitation":"(Douglas 2009, 446)","plainCitation":"(Douglas 2009, 446)"},"citationItems":[{"id":550,"uris":[""],"uri":[""],"itemData":{"id":550,"type":"article-journal","title":"Reintroducing Prediction to Explanation","container-title":"Philosophy of Science","page":"444-463","volume":"76","author":[{"family":"Douglas","given":"Heather"}],"issued":{"date-parts":[["2009"]]}},"locator":"446"}],"schema":""} (Douglas 2009, 446)She points out that Peter Lipton takes a similar stance ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"KTWwkmuV","properties":{"formattedCitation":"(Lipton 2004, 173)","plainCitation":"(Lipton 2004, 173)"},"citationItems":[{"id":352,"uris":[""],"uri":[""],"itemData":{"id":352,"type":"book","title":"Inference to the Best Explanation","publisher":"Routledge","publisher-place":"London and New York","edition":"Second edition","event-place":"London and New York","author":[{"family":"Lipton","given":"Peter"}],"issued":{"date-parts":[["2004"]]}},"locator":"173"}],"schema":""} (Lipton 2004, 173). Given the role prediction plays in debates about scientific realism, it is probably reasonable to surmise – though tricky to make a tight case – that this epistemic way of thinking about prediction is more common than the temporal, among philosophers of science.I am sceptical that the epistemic notion of prediction can be satisfactorily defined, or that it represents a stable or epistemologically significant category. I am suspicious of the implied distinction between directly accessing a fact and merely predicting it. I am also suspicious of this sort of philosophical reconstruction of everyday language and thought. Predictions seem paradigmatically to concern the future. The philosophical reconstruction asks us to lightly set this seeming aside. I need more argument than this before I will be persuaded to abandon the idea that prediction has something important to do with the future, and thus to abandon any serious philosophical investigation of that idea. Note, finally, that the epistemic notion of prediction makes it hard to count predictions as knowledge, since prediction is classified as an epistemic state short of knowledge. Whereas on the temporal notion, it seems natural to say that predictions may sometimes be known. They are just propositions about the future, and qualify as knowledge provided they are warranted in whatever way known propositions must be warranted. The epistemic stance’s bar on saying predictions can ever count as knowledge seems to me a strong reason for allowing temporal predictions alongside epistemic, since it seems that sometimes temporally predictive claims can be warranted well enough to be known (e.g. concerning the date and duration of the next solar eclipse), while temporally non-predictive claims can (obviously) be insecure (e.g. concerning the outcome of a football match that has already happened but whose score you have not yet heard).However, I do not need to pursue these doubts, nor to engage in unpromising verbal disputes about whether someone who makes an inference “that some event should have taken place, but they do not know that it did” ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"xGbKYpBn","properties":{"formattedCitation":"(Douglas 2009, 446)","plainCitation":"(Douglas 2009, 446)"},"citationItems":[{"id":550,"uris":[""],"uri":[""],"itemData":{"id":550,"type":"article-journal","title":"Reintroducing Prediction to Explanation","container-title":"Philosophy of Science","page":"444-463","volume":"76","author":[{"family":"Douglas","given":"Heather"}],"issued":{"date-parts":[["2009"]]}},"locator":"446"}],"schema":""} (Douglas 2009, 446) is really predicting, or making an inference that does not deserve that name. I need only disagree with Douglas and Lipton to the extent that they claim priority or exclusivity for the epistemic notion, and since no argument is offered for such a priority claim, we can safely disregard it. There may be two notions of prediction, epistemic and temporal. My interest is in the temporal notion of prediction, which I take it is clear enough to survive either alongside the epistemic notion, or after a rejection of the epistemic/temporal distinction.Third, prediction is similar to explanation in that it suffers from a product/process ambiguity ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"ikHfhh28","properties":{"formattedCitation":"(Ruben 1993, 16)","plainCitation":"(Ruben 1993, 16)"},"citationItems":[{"id":306,"uris":[""],"uri":[""],"itemData":{"id":306,"type":"chapter","title":"Introduction","container-title":"Explanation","collection-title":"Oxford Readings in Philosophy","publisher":"Oxford University Press","publisher-place":"Oxford","page":"1-16","event-place":"Oxford","editor":[{"family":"Ruben","given":"David-Hillel"}],"author":[{"family":"Ruben","given":"David-Hillel"}],"issued":{"date-parts":[["1993"]]}},"locator":"16"}],"schema":""} (Ruben 1993, 16). “Prediction” is ambiguous between a claim, that is, a propositional entity like “It will rain tomorrow”, and an activity, such as consulting one’s barometer in an effort to discover whether it will rain tomorrow ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"nuy3GH0y","properties":{"formattedCitation":"(Broadbent 2013, 86)","plainCitation":"(Broadbent 2013, 86)"},"citationItems":[{"id":543,"uris":[""],"uri":[""],"itemData":{"id":543,"type":"book","title":"Philosophy of Epidemiology","collection-title":"New Directions in the Philosophy of Science","publisher":"Palgrave Macmillan","publisher-place":"London and New York","event-place":"London and New York","author":[{"family":"Broadbent","given":"Alex"}],"issued":{"date-parts":[["2013"]]}},"locator":"86"}],"schema":""} (Broadbent 2013, 86).Fourth, we can distinguish two ways of thinking about predictive goodness, corresponding loosely to the claim/activity distinction. One might think of good predictions as settled by fact, by which I mean that all and only true predictions are good predictions. This is a natural way – though not, I will suggest, a helpful way – to think about prediction claims. It has no applicability to prediction activities, however, which are not propositional and do not take truth values. Thus there must be another way of thinking about predictive goodness, as not settled by fact.Together, these four distinctions allow us to structure our approach to the central question. We can say that a good prediction claim is one that is (a) true and (b) based on a good prediction activity. To endorse the factive notion of prediction claims is akin to saying that knowledge is true belief; lucky guesses are ruled in, and reasonable errors are ruled out. We can then ask what the appropriate “basing” relation is, and – more importantly – what makes a good prediction activity. We can confine our attention to temporal predictions that are made in the context of forecasting. Having arrived at a satisfactory treatment, we may want to extend it to deal with, or explain, the epistemic notion of prediction, and predictions in the context of testing.The first order of business, then, is to understand better what makes a prediction activity good. The basing relation, the goodness of prediction claims, and the possibility of extending the treatment to epistemic predictions and testing predictions will all depend on the nature of good prediction activity. The nature of good prediction activity will therefore be my focus in the remainder of this piece.Approaches to PredictionIn this section I review the most obvious conceptual tools that might be used to generate a philosophical theory of prediction, and find them all wanting.ExtrapolationI take it as axiomatic that predicting future events in the empirical world necessarily relies on inductive inference. Either the prediction activity itself will be inductive, or if it is deductive then its premises are inductively supported. The simplest way for induction to feature is for predictions to amount to simple enumerative inductions from similar past cases: in a word, extrapolation ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"dsGwvFPh","properties":{"formattedCitation":"(see e.g. Szklo and Nieto 2007, 376)","plainCitation":"(see e.g. Szklo and Nieto 2007, 376)"},"citationItems":[{"id":58,"uris":[""],"uri":[""],"itemData":{"id":58,"type":"book","title":"Epidemiology: Beyond the Basics","publisher":"Jones and Bartlett Publishers","publisher-place":"Boston, Toronto, London, Singapore","edition":"2nd","event-place":"Boston, Toronto, London, Singapore","author":[{"family":"Szklo","given":"Moyses"},{"family":"Nieto","given":"F. Javier"}],"issued":{"date-parts":[["2007"]]}},"locator":"376","prefix":"see e.g. "}],"schema":""} (see e.g. Szklo and Nieto 2007, 376).The difficulty with extrapolation as a model of good prediction is that extrapolations are often poor predictions. Share prices, for example, go down as well as up. Just because a price went up yesterday does not mean it will go up today.The natural retort is that a good extrapolation will consider the whole pattern to be extrapolated, not just a segment of it. Yesterday’s trend on a particular share price, to pursue that example, is a poor basis for extrapolation because that is not the complete body of data from which to extrapolate. There are price cycles lasting longer than a day.At this point we must distinguish between two ways one might take the question “How good is this prediction?” The first way of taking the question we might call the internal sense. Supposing for sake of illustration that we confine ourselves to linear relationships in bivariate data, the internal question is answered well by the linear correlation coefficient r, which tells us how strong a given linear relationship. The closer r is to 1 or -1, the stronger the positive or negative correlation, and thus the better any given prediction of an unobserved value of one variable based on the value of the other is likely to be.The latter claim is no logical truth. It relies on the assumption that the linear relationship in the observed data holds also for the unobserved data: that the pattern in the observed data represents the pattern in all the data, observed and unobserved. The second, external sense of the question asks whether this is true.Generalising, we can ask either the internal or the external question of a given method. We can ask how good a given prediction using that method is likely to be, given the assumptions necessary for the method to work (the internal question); or we can ask whether the circumstances are propitious for the method to work (the external question).In the case of extrapolation, “Expect more of the same” really means “expect more of the same in the same circumstances”. Bertrand Russell ably demonstrated the point by imagining a chicken who wrongly extrapolated from past experience that the approaching farmer would feed it, only to have its neck wrung ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"RIXFSBB9","properties":{"formattedCitation":"(Russell 1912 Ch 6)","plainCitation":"(Russell 1912 Ch 6)"},"citationItems":[{"id":317,"uris":[""],"uri":[""],"itemData":{"id":317,"type":"book","title":"The Problems of Philosophy","collection-title":"Great Books in Philosophy","publisher":"Prometheus Books","publisher-place":"Buffalo, New York","edition":"1988","event-place":"Buffalo, New York","author":[{"family":"Russell","given":"Bertrand"}],"issued":{"date-parts":[["1912"]]}},"suffix":"Ch 6"}],"schema":""} (Russell 1912 Ch 6). In some circumstances, the exact opposite of an extrapolation is warranted.What this shows is that the notion of extrapolation is on its own not sufficient to ground good prediction. One is right to extrapolate when circumstances are right for one to extrapolate, as well as what to do when they are not. A good prediction activity tells us what these circumstances are, and extrapolation does not. A good prediction activity may employ extrapolation, but it will answer the external question as to whether circumstances are right to do so.Prediction from Laws of NaturePerhaps a good prediction claim is one derived from a law of nature, with the “deriving” being the good prediction activity. This would yield a close symmetry with the deductive-nomological (DN) theory of explanation, and this symmetry has received considerable discussion ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"yPxNAWui","properties":{"formattedCitation":"(e.g. C. Hempel and Oppenheim 1948; Scheffler 1957; Rescher 1958; Canfield and Lehrer 1961; Rescher 1963; Kim 1964; Suchting 1967; Hanna 1969; Douglas 2009)","plainCitation":"(e.g. C. Hempel and Oppenheim 1948; Scheffler 1957; Rescher 1958; Canfield and Lehrer 1961; Rescher 1963; Kim 1964; Suchting 1967; Hanna 1969; Douglas 2009)"},"citationItems":[{"id":481,"uris":[""],"uri":[""],"itemData":{"id":481,"type":"article-journal","title":"Studies in the Logic of Explanation","container-title":"Philosophy of Science","page":"135-175","volume":"15","issue":"2","author":[{"family":"Hempel","given":"Carl"},{"family":"Oppenheim","given":"Paul"}],"issued":{"date-parts":[["1948"]]}},"prefix":"e.g."},{"id":533,"uris":[""],"uri":[""],"itemData":{"id":533,"type":"article-journal","title":"Explanation, Prediction, and Abstraction","container-title":"The British Journal for the Philosophy of Science","page":"293-309","volume":"7","issue":"28","author":[{"family":"Scheffler","given":"Israel"}],"issued":{"date-parts":[["1957"]]}}},{"id":549,"uris":[""],"uri":[""],"itemData":{"id":549,"type":"article-journal","title":"On Prediction and Explanation","page":"281-290","volume":"8","issue":"32","author":[{"family":"Rescher","given":"Nicholas"}],"issued":{"date-parts":[["1958"]]}}},{"id":559,"uris":[""],"uri":[""],"itemData":{"id":559,"type":"article-journal","title":"A Note on Prediction and Deduction","container-title":"Philosophy of Science","page":"204-208","volume":"28","issue":"2","author":[{"family":"Canfield","given":"John"},{"family":"Lehrer","given":"Keith"}],"issued":{"date-parts":[["1961"]]}}},{"id":530,"uris":[""],"uri":[""],"itemData":{"id":530,"type":"article-journal","title":"Discrete State Systems, Markov Chains, and Problems in the Theory of Scientific Explanation and Prediction","container-title":"Philosophy of Science","page":"325-345","volume":"30","issue":"4","author":[{"family":"Rescher","given":"Nicholas"}],"issued":{"date-parts":[["1963"]]}}},{"id":541,"uris":[""],"uri":[""],"itemData":{"id":541,"type":"article-journal","title":"Inference, Explanation, and Prediction","container-title":"The Journal of Philosophy","page":"360-368","volume":"61","issue":"12","author":[{"family":"Kim","given":"Jaegwon"}],"issued":{"date-parts":[["1964"]]}}},{"id":544,"uris":[""],"uri":[""],"itemData":{"id":544,"type":"article-journal","title":"Deductive Explanation and Prediction Revisited","container-title":"Philosophy of Science","page":"41-52","volume":"34","issue":"1","author":[{"family":"Suchting","given":"WA"}],"issued":{"date-parts":[["1967"]]}}},{"id":528,"uris":[""],"uri":[""],"itemData":{"id":528,"type":"article-journal","title":"Explanation, Prediction, Description, and Information Theory","container-title":"Synthese","page":"308-334","volume":"20","issue":"3","author":[{"family":"Hanna","given":"Joseph F."}],"issued":{"date-parts":[["1969"]]}}},{"id":550,"uris":[""],"uri":[""],"itemData":{"id":550,"type":"article-journal","title":"Reintroducing Prediction to Explanation","container-title":"Philosophy of Science","page":"444-463","volume":"76","author":[{"family":"Douglas","given":"Heather"}],"issued":{"date-parts":[["2009"]]}}}],"schema":""} (e.g. C. Hempel and Oppenheim 1948; Scheffler 1957; Rescher 1958; Canfield and Lehrer 1961; Rescher 1963; Kim 1964; Suchting 1967; Hanna 1969; Douglas 2009). Is there a viable DN theory of prediction?It is hard to dispute the claim that laws of nature may be useful for making predictions. If one wishes to predict the next solar eclipse, DN prediction is likely to work very well. But the solar system is a special case, as perceptive commentators have noted ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"W2g3tt1w","properties":{"formattedCitation":"(Anscombe 1971)","plainCitation":"(Anscombe 1971)"},"citationItems":[{"id":293,"uris":[""],"uri":[""],"itemData":{"id":293,"type":"book","title":"Causality and Determination","publisher":"Cambridge University Press","publisher-place":"Cambridge","event-place":"Cambridge","author":[{"family":"Anscombe","given":"G. E.M"}],"issued":{"date-parts":[["1971"]]}}}],"schema":""} (Anscombe 1971), since it is a system that proceeds in an orderly fashion with remarkably little intervention from external factors within human timeframes.The central difficulty with DN prediction is that it cannot handle intervention by external factors, of the kind that are largely absent from the progress of the solar system over the timeframe of a prediction of the next solar eclipse. All known scientific laws only hold ceteris paribus – “other things being equal”. As for extrapolation, the question then becomes whether other things are equal. The external question for a prediction relying on laws is whether the ceteris paribus clause is satisfied, and whether the initial conditions are as one supposes them to be. DN prediction does not suffice for good prediction, because the mere fact that a claim is deduced from laws plus initial conditions does not suffice to show that the ceteris paribus clause for that law is satisfied. Even when predicting an eclipse, the goodness of a prediction invoking laws depends upon our background knowledge that intrusions into the Solar System are highly unlikely, as well as on assumptions about the negligibility of inevitable measurement errors concerning the initial positions of the planets.These assumptions might be perfectly legitimate. Again, I am not disputing that deduction from natural law is a reasonable way to predict solar eclipses. Rather, I am disputing the idea that deduction from law provides the basis of a theory of good prediction. A theory of good prediction must tell us why this is a good way of predicting eclipses. The contemplated answer is, “Because it uses laws of nature.” That is a bad answer because it does not distinguish this case from other circumstances, such as predicting the weather, where deduction from laws of nature is a bad prediction activity. There is thus more to good prediction than deduction from natural law.Prediction from CausesIt is commonly supposed among those who think about public policy that knowledge of causes is necessary for reliably predicting the outcome of policy interventions ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"hTA1xr32","properties":{"formattedCitation":"(see e.g. Rutter 2007)","plainCitation":"(see e.g. Rutter 2007)"},"citationItems":[{"id":96,"uris":[""],"uri":[""],"itemData":{"id":96,"type":"report","title":"Identifying the environmental causes of disease: how should we decide what to believe and when to take action?","publisher":"The Academy of Medical Sciences","author":[{"family":"Rutter","given":"Michael"}],"issued":{"date-parts":[["2007"]]}},"prefix":"see e.g."}],"schema":""} (see e.g. Rutter 2007). Causal knowledge can protect us from being led into error when mere extrapolation goes wrong. If, for example, we know that a program to educate mothers about infant nutrition worked in part because mothers in that region control food purchases and distribution within the household, then we will not mistakenly predict that it will work in a region where husbands do the shopping and mothers-in-law head the household ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"nEQ5dPKw","properties":{"formattedCitation":"(Cartwright 2010)","plainCitation":"(Cartwright 2010)"},"citationItems":[{"id":301,"uris":[""],"uri":[""],"itemData":{"id":301,"type":"paper-conference","title":"Will this policy work for you? Predicting effectiveness better: How philosophy helps","container-title":"PSA Presidential Address","event":"Philosophy of Science Association Biennial Meeting","URL":"","author":[{"family":"Cartwright","given":"Nancy"}],"issued":{"date-parts":[["2010"]]},"accessed":{"date-parts":[["2012",5,1]]}}}],"schema":""} (Cartwright 2010). Perhaps there are classes of predictions that may count as good despite a lack of causal knowledge (I invite the reader to identify some), but at least as practical a rule of thumb, it seems safe to exclude as good those predictions where the causes of the predicted phenomena either are not known or do not form part of the justification for the prediction.However, causal knowledge is clearly not sufficient for prediction, and as such we cannot rest with a simple causal model of good prediction. There are some thorny issues here concerning the distinction between general and particular causal claims. If we know the truth of a particular causal claim then (standardly) that implies the occurrence of the effect in question. But this is not the case in the context of forecasting. When forecasting, we usually know that a given event can cause a given effect. But we may not know that it will cause that effect on this occasion. The same event can be a cause of both E and not-E. For example, if housing prices crash tomorrow, the rate of inflation will likely be among the causes. But if housing prices rise tomorrow, the rate of inflation will also be among the causes. It contributes in either case. If there is a storm tomorrow, the air pressure is among the causes; and likewise if there is not. If you speak your mind in a meeting, the presence of oxygen is among the causes; and likewise if you remain silent. Many causes of actual effects are like this: they could also have been among causes of other events that did not in fact occur. From the point of view of predicting, this means that merely because a given event can cause something does not mean that it will on a particular occasion, and therefore that some warrant is required for the inference that a cause that is capable of having a given effect will in fact have that effect.We need to know how to move from “can cause” to “will cause”. This move will depend on knowledge of other causal circumstances. Whether oxygen causes silence or speech depends on what else is going on. We might say that causal knowledge is sufficient for prediction when the knowledge is of a sufficient cause; but that is a bad move, because we probably never have knowledge of sufficient causes in their entirety – such knowledge would need to exclude every possible intervener, including meteorites from out of space, solar flares, highly unlikely quantum events, and so forth. If good prediction requires knowledge of that sort then we can never achieve it. Surely the difficulty of knowing these things does have something to do with the difficulty of prediction. To invoke sufficient causes is tantamount to saying that warrant for using causes to predict only comes when one knows all the circumstances are as they need to be. Perhaps; but that would be depressing if so. If we want to draw a useful line between good and bad predictions, we need a standard that can be humanly satisfied, and so should reject those standards that clearly would yield true predictions but are equally clearly unattainable by humans, just as we reject those standards that would wrongly judge bad predictions to be good.Prediction from MechanismMuch, probably too much, has been made of the notion of mechanism in recent philosophy of science. Perhaps prediction requires knowledge of an underlying mechanism. Such knowledge is presumably a kind of causal knowledge, but with bells and whistles: one does not simply know that A causes B; one has a detailed picture of how A causes B. It is the difference between knowing that winding your watch causes the hands to go round, and understanding how winding causes the hands to go round – understanding the watch mechanism. And indeed it is plausible that someone who understands the watch mechanism will make better predictions about the watch than someone who merely knows that winding causes ticking.Knowledge of mechanisms, as a kind of detailed causal knowledge, is surely useful for predicting. But it is neither necessary nor sufficient, as I have argued elsewhere [author’s ref]. Prof Barometer knows how his barometer works but is ignorant of the fact that air pressure falls with altitude, due to his cloistered life at a university located in a low-lying English swamp. His predictions about the weather go wrong when he visits his old school friend Joe User, who lives in a hilly region. Joe is ignorant about the workings of barometers but knows from experience that changes in altitude affect their reliability; he is safe from a set of errors that Prof Barometer is not, and is in fact a more reliable predictor using his barometer. Prof Barometer’s knowledge of the mechanism is not sufficient for good prediction. And Joe User’s ignorance of the mechanism does not prevent him from making good predictions, showing that knowledge of the mechanism is not necessary for good prediction.Again, circumstances matter. One can predict using mechanistic knowledge, provided that the circumstances are right for the mechanism to work. In different circumstances the same mechanistic knowledge might yield quite different predictions. Our question, “How good is this prediction?”, taken in the external sense, asks what warrant there is for supposing that the circumstances are right for the mechanism to work and not break. Although further mechanistic knowledge may be invoked in establishing this warrant, the mere fact that a mechanism is used to make a prediction obviously does not in itself establish that circumstances were right for its use. Thus our theory of good prediction cannot start from the point that a good prediction is one that employs mechanistic knowledge, since bad predictions also can.The Psychology of Good PredictionThe psychologist Philip Tetlock has conducted a long term study of the predictive accuracy of putative experts in the social and political sciences ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"W1Lp3EFs","properties":{"formattedCitation":"(Tetlock 2005)","plainCitation":"(Tetlock 2005)"},"citationItems":[{"id":525,"uris":[""],"uri":[""],"itemData":{"id":525,"type":"book","title":"Expert Political Judgement - How Good Is It? How Can We Know?","publisher":"Princeton University Press","publisher-place":"Princeton and Oxford","event-place":"Princeton and Oxford","author":[{"family":"Tetlock","given":"Philip"}],"issued":{"date-parts":[["2005"]]}}}],"schema":""} (Tetlock 2005). “Putative” because their accuracy is poor enough to cast doubt on their expert status in many cases. Predicting political events such as the fall of the Soviet Union may be different from predicting natural events such as tomorrow’s weather. But it would be foolish to ignore this evidence merely on that basis.Tetlock posed “large numbers of experts large numbers of questions about large numbers of cases” and applied “no-favoritism scoring rules to the answers” ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"z0D93aZn","properties":{"formattedCitation":"(Tetlock 2005, 8)","plainCitation":"(Tetlock 2005, 8)"},"citationItems":[{"id":525,"uris":[""],"uri":[""],"itemData":{"id":525,"type":"book","title":"Expert Political Judgement - How Good Is It? How Can We Know?","publisher":"Princeton University Press","publisher-place":"Princeton and Oxford","event-place":"Princeton and Oxford","author":[{"family":"Tetlock","given":"Philip"}],"issued":{"date-parts":[["2005"]]}},"locator":"8"}],"schema":""} (Tetlock 2005, 8). The results are sobering:When we pit experts against minimalist performance benchmarks – dilettantes, dart-throwing chimps, and assorted extrapolation algorithms – we ?nd few signs that expertise translates into greater ability to make either “well-calibrated” or “discriminating” forecasts. ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"8WIpMPhd","properties":{"formattedCitation":"(Tetlock 2005, 20)","plainCitation":"(Tetlock 2005, 20)"},"citationItems":[{"id":525,"uris":[""],"uri":[""],"itemData":{"id":525,"type":"book","title":"Expert Political Judgement - How Good Is It? How Can We Know?","publisher":"Princeton University Press","publisher-place":"Princeton and Oxford","event-place":"Princeton and Oxford","author":[{"family":"Tetlock","given":"Philip"}],"issued":{"date-parts":[["2005"]]}},"locator":"20"}],"schema":""} (Tetlock 2005, 20)Perhaps even more worryingly, professional background is not significantly correlated with forecasting accuracy, including whether or not the expert has a doctorate, years of professional experience, and whether the expert has access to classified information ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"Lbpzb8Ii","properties":{"formattedCitation":"(Tetlock 2005, 69)","plainCitation":"(Tetlock 2005, 69)"},"citationItems":[{"id":525,"uris":[""],"uri":[""],"itemData":{"id":525,"type":"book","title":"Expert Political Judgement - How Good Is It? How Can We Know?","publisher":"Princeton University Press","publisher-place":"Princeton and Oxford","event-place":"Princeton and Oxford","author":[{"family":"Tetlock","given":"Philip"}],"issued":{"date-parts":[["2005"]]}},"locator":"69"}],"schema":""} (Tetlock 2005, 69).However, Tetlock does not recommend general scepticism about human predictive ability. He argues that certain cognitive styles perform noticeably better than others: they are correlated with forecasting accuracy. Tetlock takes over Isiah Berlin’s distinction between thinkers who are hedgehogs and thinkers who are foxes ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"a21frTAp","properties":{"formattedCitation":"(Berlin 2013)","plainCitation":"(Berlin 2013)"},"citationItems":[{"id":562,"uris":[""],"uri":[""],"itemData":{"id":562,"type":"book","title":"The Hedgehog and the Fox: An Essay on Tolstoy's View of History","publisher":"Princeton University Press","number-of-pages":"160","edition":"2","source":"","ISBN":"069115600X","shortTitle":"The Hedgehog and the Fox","author":[{"family":"Berlin","given":"Isaiah"}],"editor":[{"family":"Hardy","given":"Henry"}],"issued":{"date-parts":[["2013",4,21]]}}}],"schema":""} (Berlin 2013). Hedgehogs know one big thing: they have an overarching idea or intellectual tool which they use to explain and to predict everything. They tend to be intellectually aggressive, making bold claims, and resisting altering those claims when evidence comes to light suggesting that they might be wrong. They tend to be confident of their opinions, including predictive ones, presumably because they are convinced of the truth of the big idea that generates them. Foxes, on the other hand, use a range of cognitive tools, respond to changing evidential situations, qualify their predictions, and are less certain of them. Tetlock argues that foxy thinking encourages ideological moderation, while hedgehog-ish thinking lends itself to extremism ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"L2q3FlW5","properties":{"formattedCitation":"(Tetlock 2005, 86)","plainCitation":"(Tetlock 2005, 86)"},"citationItems":[{"id":525,"uris":[""],"uri":[""],"itemData":{"id":525,"type":"book","title":"Expert Political Judgement - How Good Is It? How Can We Know?","publisher":"Princeton University Press","publisher-place":"Princeton and Oxford","event-place":"Princeton and Oxford","author":[{"family":"Tetlock","given":"Philip"}],"issued":{"date-parts":[["2005"]]}},"locator":"86"}],"schema":""} (Tetlock 2005, 86). Elsewhere I have advocated a simple heuristic for prediction by asking the question, “What could possibly go wrong?” I have not done Tetlock’s study justice in this summary, but it nevertheless provides some empirical plausibility that thinkers who follow this rule of thumb are better predictors than those who do not. This suggests that it would not be entirely hedgehog-ish to seek to develop and generalise an account of prediction along lines which I have gestured at elsewhere [author’s ref].A Theory of Good PredictionThe trouble with seeking a model of prediction in terms of any of the concepts identified in Section REF _Ref349132450 \r \h \* MERGEFORMAT III is not that they are useless for making predictions. On the contrary, without appealing to one or other of these notions, it is hard to see how prediction could amount to more than guesswork. Taken together, the review of conceptual resources in Section REF _Ref349132450 \r \h III and psychological evidence in Section REF _Ref349132476 \r \h IV suggests that a good prediction activity is not so much a matter of which method you pick, but of ensuring that circumstances are right for your chosen method to work.Thus the task of coming up with a philosophical theory of prediction is narrowed to the slightly less daunting task of coming up with an account of the rightness of circumstances for a predictive method to be employed. To make the task less daunting still, I want to draw an analogy between prediction and the standard realist picture of the way confirmation works for claims about unobservables. (This analogy does not commit us to scientific realism, only to realism about the future, which I take to be acceptable to the empiricist and realist alike.) The hope is that the way realists think we can get knowledge of unobservables provides a good starting point for explaining how we might get knowledge of the future.Suppose a scientist entertains some claim T whose truth is not readily observable (whatever that exactly means). T might be a definite hypothesis, for example, that there are subatomic particles, known as electrons. These particles cannot be seen, nor detected with any existing instrument. How does the scientist test this theory and gather confirmatory or disconfirmatory evidence? On a standard realist view, she works out what observable differences there are between her hypothesis and various incompatible alternative hypotheses, and then she makes observations concerning those differences. Her theory is confirmed if the observable differences are in favour of her theory, and disconfirmed otherwise.In this example, one aspect of electron theory is that charge is quantized, arising (on electron theory) from the distribution of charged subatomic particles. If it is possible to measure the charge on an electron, then one will have evidence that tells between the hypothesis that charge is quantized and that it is continuous, and so between the hypothesis that there are subatomic particles, including electrons, and that there are not. Thus when Robert Millikan experimentally measured the charge on the electron, he provided evidence that simultaneously confirmed one hypothesis and disconfirmed another. Millikan’s experiment would have been of much less interest if its results had been equally compatible with either theory. Let me stress two features of this picture. First, not just any incompatible hypothesis is countenanced as providing a useful foil for the theory under test. The foil must have something going for it, in order for beating that hypothesis to count in favour of the theoretical claim in question. One does not (as in a cartoon I have seen) strap puppies to missiles on the basis that good science requires testing our assumptions about the likely outcome. This is not because we have a wealth of data about puppies that have been fired off with missiles in the past; it is rather because there is no serious rival to the obvious account of their fate. Impressive confirmatory evidence comes from tests that distinguish a theory from serious rivals, not from just any contrary empirical claim.Second, contrasts play an important role. This way (and maybe there are others) of obtaining evidence about unobservables seems to depend minimally upon identifying differences between how things would be if the theory is true, and how things would be if the theory were false. Note in particular that “contrast” does not just refer to the negation of the hypothesis under test. Substantive contrasts, i.e. contrasts between T and something logically stronger than not-T, are important because the mere negation of a hypothesis about unobservables does not necessarily imply any observable difference. Denying that there electrons does not, by itself, yield the consequence that charge is not quantized. To generate a rival observable claim, one needs a rival theory that is logically strong enough for the job. Merely negating a theory about unobservables does not usually yield any observable consequences at all. The two points are connected because the negation of a serious hypothesis is not necessarily, and indeed not usually, a serious rival. We don’t strap puppies to missiles to test the hypothesis that they will be blown to smithereens against its mere negation, since the mere negation is not a serious rival hypothesis, even though the hypothesis it negates is highly credible. The point is reminiscent of Quine’s claim that the complement of a projectible predicate is not necessarily projectible ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"PsDUeVbR","properties":{"formattedCitation":"(Quine 1969, 115)","plainCitation":"(Quine 1969, 115)"},"citationItems":[{"id":129,"uris":[""],"uri":[""],"itemData":{"id":129,"type":"chapter","title":"Natural Kinds","container-title":"Ontological Relativity and Other Essays","publisher":"Columbia University Press","publisher-place":"New York","page":"114―138","event-place":"New York","author":[{"family":"Quine","given":"W. V.O"}],"issued":{"date-parts":[["1969"]]}},"locator":"115"}],"schema":""} (Quine 1969, 115).I want to propose a theory of good prediction activity that is analogous to the manner in which claims about unobservables are, in the realist’s eyes, confirmed by observations. A good prediction activity is one that (a) identifies and (b) verifies currently observable consequences of the prediction claim it supports.“Currently observable consequences” satisfy two conditions. First, the predictor is able to observe them before the date that the prediction concerns. They are analogous to the “observable consequences” of unobservable claims, although clearly any causal implication in “consequence” will need to be rethought for the case of prediction, where the matters that one infers lie in the future of one’s evidence. We will discuss this shortly.Second, currently observable consequences of the truth of the prediction are also differences from currently observable consequences of some foil that is contrary to the prediction. The two remarks concerning contrastive inference apply as follows. First, defeating the foil confers support on the prediction only to the extent that the foil enjoys support of its own. Second, some substantive, contentful foil, and not merely the denial of the prediction claim, is necessary in order to be able to generate the requisite currently observable consequences against which to test the prediction claim.Confusion might arise at this stage because the analogy with theoretical claims about unobservables suggests that it is the prediction claim that is being tested, whereas I have so far talked about checking that circumstances are right for the method that was employed to produce the prediction. The confusion is resolved by seeing inappropriate circumstances for a given method as one way that a predictive claim might be cast into doubt. The truth of the prediction claim is the ultimate goal, so it makes sense to focus on testing the claim against rival ways things might go. In my critique, I have focused in particular on the failure of putative methods for arriving at a prediction, because I have been concerned to refute the idea that employing any of these methods is sufficient for good prediction. But there is no reason to confine our assessment of prediction claims to an assessment of the methods used to produce them.The analogy between predictions and confirmation of theories about unobservables gets us only so far. It breaks down when we consider the causal structures underlying this kind of inference in each case. The realist picture of theoretical confirmation is often cast as inference to the best explanation. The theory T explains experimental results better than rival theories, and, according to the proponent of inference to the best explanation, is therefore not merely preferable to those other theories, but likely to be true, if the rival theories are good enough ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"dNbU3Uue","properties":{"formattedCitation":"(Lipton 2004 esp. Chs 3-5)","plainCitation":"(Lipton 2004 esp. Chs 3-5)"},"citationItems":[{"id":352,"uris":[""],"uri":[""],"itemData":{"id":352,"type":"book","title":"Inference to the Best Explanation","publisher":"Routledge","publisher-place":"London and New York","edition":"Second edition","event-place":"London and New York","author":[{"family":"Lipton","given":"Peter"}],"issued":{"date-parts":[["2004"]]}},"suffix":"esp. Chs 3-5"}],"schema":""} (Lipton 2004 esp. Chs 3-5). But future events cannot explain past ones, so we cannot cast predictive inference as inference to the best explanation, without further elaboration.In the case of prediction, explanation is still involved, but the explanans is not the prediction claim. On the model I have sketched, the business of making a prediction good is one of explaining why it rather than alternatives is true. The alternatives are those predictive claims entailing different currently observable consequences. In verifying the currently observable consequences, one simultaneously rules out these rival theories that contradict the prediction claim you are supporting. You are thus giving causal differences between the situation that will obtain if the prediction is true, and that which will obtain if various alternatives are true. In other words, you are supplying a contrastive causal explanation of the contrast between the event predicted and some foil.This model resolves a longstanding confusion about the relation between prediction and explanation. On the one hand, there clearly seems to be a link. Douglas argues that explanation provides cognitive tools for making predictions ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"emwc1XiB","properties":{"formattedCitation":"(Douglas 2009)","plainCitation":"(Douglas 2009)"},"citationItems":[{"id":550,"uris":[""],"uri":[""],"itemData":{"id":550,"type":"article-journal","title":"Reintroducing Prediction to Explanation","container-title":"Philosophy of Science","page":"444-463","volume":"76","author":[{"family":"Douglas","given":"Heather"}],"issued":{"date-parts":[["2009"]]}}}],"schema":""} (Douglas 2009). The theory proposed here is consonant with that view and puts some flesh on the bones, by showing how explanation is involved in securing the warrant for a prediction. On the other hand, counterexamples to the claim that every prediction is an explanation abound. Some of these are handled by the distinction between prediction claims and prediction activities. Merely making a claim about the future clearly does not require explanation. Others are handled by the notion of good prediction. A prediction activity may not involve explanation, but on this model, a good predictive activity will include a series of contrastive explanations as to why the prediction claim, and not any plausible contrary, is true.ConclusionIn the context of forecasting, a good prediction claim is a true claim about the future based on a good prediction activity, and a good prediction activity is one which identifies and verifies observable consequences of the prediction claim it supports. That is the full statement of the theory of prediction that we have arrived at, and of the answer to the title question of this paper.There remain a number of loose ends and unanswered questions. First, I have not attempted to demarcate “scientific prediction”, in the manner that Hempel sought to demarcate scientific explanation. Second, although I have sought to connect good prediction with other epistemic exercises, and with forms of inference that are thought by some to yield knowledge (I am referring to the analogy with confirmation of claims about unobservables), I have not squarely addressed the question whether predictions are ever knowledge of the future. Third, I have not tried to decide whether only an internalist account can provide a model of good prediction, or whether the references to checking, explanation, and so forth can be schematized in an externalist fashion. My hunch is that it can’t be done: that good prediction, as I have characterised it, is necessarily an active intellectual exercise. But I have not argued this case.References ADDIN ZOTERO_BIBL {"custom":[]} CSL_BIBLIOGRAPHY Anscombe, G. E.M. 1971. Causality and Determination. Cambridge: Cambridge University Press.Berlin, Isaiah. 2013. The Hedgehog and the Fox: An Essay on Tolstoy’s View of History. Edited by Henry Hardy. 2nd ed. Princeton University Press.Broadbent, Alex. 2013. Philosophy of Epidemiology. New Directions in the Philosophy of Science. London and New York: Palgrave Macmillan.Canfield, John, and Keith Lehrer. 1961. “A Note on Prediction and Deduction.” Philosophy of Science 28 (2): 204–208.Cartwright, Nancy. 2010. “Will This Policy Work for You? Predicting Effectiveness Better: How Philosophy Helps.” In PSA Presidential Address. .———. 2011. “Predicting What Will Happen When We Act. What Counts for Warrant?” Preventive Medicine 53 (4-5): 221–4.Douglas, Heather. 2009. “Reintroducing Prediction to Explanation.” Philosophy of Science 76: 444–463.Duhem, Pierre. 1914. The Aim and Structure of Physical Theory. New York: Atheneum.Goodman, Nelson. 1983. Fact, Fiction and Forecast. Fourth. Cambridge, Massachusetts: Harvard University Press.Hall, Ned. 2004. “Causation and the Price of Transitivity.” In Causation and Counterfactuals, edited by J. Collins, N. Hall, and L. A Paul, 181―204. Cambridge, Massachusetts: MIT Press.Halpern, Joseph Y., and Judea Pearl. 2005. “Causes and Explanations: A Structural Models Approach. Part 1: Causes.” British Journal for the Philosophy of Science 56: 843–887.Hanna, Joseph F. 1969. “Explanation, Prediction, Description, and Information Theory.” Synthese 20 (3): 308–334.Hempel, Carl G. 1945a. “Studies in the Logic of Confirmation (I.).” Mind 54: 1–26.———. 1945b. “Studies in the Logic of Confirmation (II.).” Mind 54: 97–121.Hempel, Carl, and Paul Oppenheim. 1948. “Studies in the Logic of Explanation.” Philosophy of Science 15 (2): 135–175.Kim, Jaegwon. 1964. “Inference, Explanation, and Prediction.” The Journal of Philosophy 61 (12): 360–368.Laudan, Larry. 1981. “A Confutation of Convergent Realism.” Philosophy of Science 48 (1): 19–49.Lawson, Tony. 1985. “The Context of Prediction (and the Paradox of Confirmation).” The British Journal for the Philosophy of Science 36 (4): 393–407.Lewis, David. 1973. “Causation.” The Journal of Philosophy 70 (17): 556–567.Lipton, Peter. 2004. Inference to the Best Explanation. Second edition. London and New York: Routledge.———. 2005. “Testing Hypotheses: Prediction and Prejudice.” Science 307 (5707): 219–221.Maher, Patrick. 1988. “Prediction, Accommodation, and the Logic of Discovery.” PSA: Proceedings of the Biennial Meeting of the Philosophy of Science Association 1: 273–285.———. 1993. “Howson and Franklin on Prediction.” Philosophy of Science 60 (2): 329–340.Matthew, Anthony. 1971. “Prediction and Predication.” The British Journal for the Philosophy of Science 22 (2): 171–182.Merrill, G.H. 1979. “Prediction and Confirmation.” Philosophy of Science 46 (1): 98–117.Pearl, Judea. 2009. Causality: Models, Reasoning and Inference. 2nd ed. Cambridge: Cambridge University Press.Quine, W. V.O. 1953. “Two Dogmas of Empiricism.” In From a Logical Point of View, 20―43. Cambridge, Mass.: Harvard University Press.———. 1969. “Natural Kinds.” In Ontological Relativity and Other Essays, 114―138. New York: Columbia University Press.Rescher, Nicholas. 1958. “On Prediction and Explanation” 8 (32): 281–290.———. 1963. “Discrete State Systems, Markov Chains, and Problems in the Theory of Scientific Explanation and Prediction.” Philosophy of Science 30 (4): 325–345.Ruben, David-Hillel. 1993. “Introduction.” In Explanation, edited by David-Hillel Ruben, 1–16. Oxford Readings in Philosophy. Oxford: Oxford University Press.Russell, Bertrand. 1912. The Problems of Philosophy. 1988th ed. Great Books in Philosophy. Buffalo, New York: Prometheus Books.Rutter, Michael. 2007. “Identifying the Environmental Causes of Disease: How Should We Decide What to Believe and When to Take Action?” The Academy of Medical Sciences.Salmon, Wesley C. 1981. “Rational Prediction.” The British Journal for the Philosophy of Science 32 (2): 115–125.Schaffer, Jonathan. 2005. “Contrastive Causation.” Philosophical Review 114 (3): 297–328.———. 2007. “The Metaphysics of Causation.” Stanford Encyclopedia of Philosophy. .———. 2010. “Contrastive Causation in the Law.” Legal Theory 16: 259–297.Scheffler, Israel. 1957. “Explanation, Prediction, and Abstraction.” The British Journal for the Philosophy of Science 7 (28): 293–309.Spirtes, P., C. Glymour, and R. Scheines. 2000. Causation, Prediction, and Search. 2nd ed. New York, N.Y.: Springer-Verlag.Suchting, WA. 1967. “Deductive Explanation and Prediction Revisited.” Philosophy of Science 34 (1): 41–52.Szklo, Moyses, and F. Javier Nieto. 2007. Epidemiology: Beyond the Basics. 2nd ed. Boston, Toronto, London, Singapore: Jones and Bartlett Publishers.Tetlock, Philip. 2005. Expert Political Judgement - How Good Is It? How Can We Know? Princeton and Oxford: Princeton University Press.Van Fraassen, Bas. 1980. The Scientific Image. Oxford: Oxford University Press.White, Roger. 2003. “The Epistemic Advantage of Prediction over Accommodation.” Mind 112 (448): 653–683.Whitmore, Charles E. 1942. “The Status of Prediction.” The Journal of Philosophy 39 (1): 5–14. ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download