develops themes from the dissertation. I argue that two models of prosopagnosia are best understood as idealizing models, and as such are subject to importantly different methodological constraints from non-idealized theories of face recognition.
In a target article that appeared in this journal, Thomas Stoffregen 2000 questions the possibility of ecological event perception research. This paper describes an experiments performed to examine the perception of the disappearance of gap-crossing affordances, a variety of event as defined by Chemero 2000. We found that subjects reliably perceive both gap-crossing affordances and the disappearance of gap-crossing affordances. Our findings provide empirical evidence in favor of understanding events as changes in the layout of affordances, shoring up event perception (...) research in ecological psychology. (shrink)
Reasonable individuals can disagree about philosophical questions. This disagreement sometimes takes the form of conflicting intuitions; the seminar room provides many examples. Experimental philosophers, who have devoted themselves to the systematic study of intuitions, have found empirical support for what anecdotes suggest. Their data often reveals that a significant minority of subjects have intuitions counter to those of the majority.1 A recent replication of [Knobe, 2003a] discovered three distinct subgroups of subjects with three distinct patterns of response. Only about one-third (...) of subjects gave the expected asymmetric responses to Knobestyle probes. The other two-thirds gave symmetric responses to harm and help probes, with about half judging side-effects were intentional in both harm and help cases and half judging that they were intentional in neither [Nichols and Ulatowski, 2007]. (shrink)
The problem with idealization is not just that, when idealizing, scientists ask us to suppose false things. Many people do that. No, the puzzling thing about idealizers—unlike astrologers, spodomancers, and homeopaths—is that it is worth listening to them. Supposing that populations of rabbits are in- finite is useful for a variety of ecological explanations. Yet we are not up to our necks in rabbits; the puzzle is why it should be useful to suppose that we are.
Are spheres multiply realizable? A venerable tradition implies that they are. Putnam’s discussion of the peg and holes (in [Putnam, 1975]) is often taken to show that all volumetric shape properties are multiply realizable . The argument runs: (a) physics is the science of the “ultimate constituents” (Putnam’s phrase) of matter, and so (b) physics can only track the behavior of each of the simple constituents of a particular system, but (c) tediously tracking individual particles doesn’t make for a very (...) good explanation, so (d) there must be an explanation outside of physics that does talk about shape, and that we should prefer because it abstracts away from the micro-level details of particular spheres. (shrink)
Functional magnetic resonance imaging (or fMRI)1 is widely used to support hypotheses about brain function. Many find the images produced from fMRI data to be especially compelling evidence for scientific hypotheses [McCabe and Castel, 2008]. There are many problems with all of this; I want to start with two of them, and argue that they get us closer to an under-appreciated worry about many imaging experiments.
Pains motivate us. Must they? Motivationalists about pain say yes: motivational force is an intrinsic property of pains. Many disagree. The debate is partly empirical. Find someone who is entirely unmoved by pain, and motivationalism is threatened. Fail repeatedly to find such a case, and motivationalism gains credence.
We discuss two named-entity recognition models which use characters and character n-grams either exclusively or as an important part of their data representation. The ﬁrst model is a character-level HMM with minimal context information, and the second model is a maximum-entropy conditional markov model with substantially richer context features. Our best model achieves an overall F1 of 86.07% on the English test data (92.31% on the development data). This number represents a 25% error reduction over the same model without word-internal (...) (substring) features. (shrink)
Method by subtracting off the error along several nonprincipal eigenvectors from the current iterate of the Power Method, making use of known nonprincipal eigenvalues of the Web hyperlink matrix. Empirically, we show that using Power Extrapolation speeds up PageRank computation by 30% on a Web graph of 80 million nodes in realistic scenarios over the standard power method, in a way that is simple to understand and implement.
While Ç ´Ò¿ µ methods for parsing probabilistic context-free grammars (PCFGs) are well known, a tabular parsing framework for arbitrary PCFGs which allows for botton-up, topdown, and other parsing strategies, has not yet been provided. This paper presents such an algorithm, and shows its correctness and advantages over prior work. The paper ﬁnishes by bringing out the connections between the algorithm and work on hypergraphs, which permits us to extend the presented Viterbi (best parse) algorithm to an inside (total probability) (...) algorithm. (shrink)
We present a generative distributional model for the unsupervised induction of natural language syntax which explicitly models constituent yields and contexts. Parameter search with EM produces higher quality analyses than previously exhibited by unsupervised systems, giving the best published unsupervised parsing results on the ATIS corpus. Experiments on Penn treebank sentences of comparable length show an even higher F1 of 71% on nontrivial brackets. We compare distributionally induced and actual part-of-speech tags as input data, and examine extensions to the basic (...) model. We discuss errors made by the system, compare the system to previous models, and discuss upper bounds, lower bounds, and stability for this task. (shrink)
A* PCFG parsing can dramatically reduce the time required to find the exact Viterbi parse by conservatively estimating outside Viterbi probabilities. We discuss various estimates and give efficient algorithms for computing them. On Penn treebank sentences, our most detailed estimate reduces the total number of edges processed to less than 3% of that required by exhaustive parsing, and even a simpler estimate which can be pre-computed in under a minute still reduces the work by a factor of 5. The algorithm (...) extends the classic A* graph search procedure to a certain hypergraph associated with parsing. Unlike bestfirst and finite-beam methods for achieving this kind of speed-up, the A* parser is guaranteed to return the most likely parse, not just an approximation. The algorithm is also correct for a wide range of parser control strategies and maintains a worst-case cubic time bound. (shrink)
We demonstrate that an unlexicalized PCFG can parse much more accurately than previously shown, by making use of simple, linguistically motivated state splits, which break down false independence assumptions latent in a vanilla treebank grammar. Indeed, its performance of 86.36% (LP/LR F1) is better than that of early lexicalized PCFG models, and surprisingly close to the current state-of-theart. This result has potential uses beyond establishing a strong lower bound on the maximum possible accuracy of unlexicalized models: an unlexicalized PCFG is (...) much more compact, easier to replicate, and easier to interpret than more complex lexical models, and the parsing algorithms are simpler, more widely understood, of lower asymptotic complexity, and easier to optimize. (shrink)
This paper separates conditional parameter estima- tion, which consistently raises test set accuracy on statistical NLP tasks, from conditional model struc- tures, such as the conditional Markov model used for maximum-entropy tagging, which tend to lower accuracy. Error analysis on part-of-speech tagging shows that the actual tagging errors made by the conditionally structured model derive not only from label bias, but also from other ways in which the independence assumptions of the conditional model structure are unsuited to linguistic sequences. The (...) paper presents new word-sense disambiguation and POS tagging experiments, and integrates apparently conﬂicting reports from other recent work. (shrink)
Unsupervised grammar induction systems commonly judge potential constituents on the basis of their effects on the likelihood of the data. Linguistic justiﬁcations of constituency, on the other hand, rely on notions such as substitutability and varying external contexts. We describe two systems for distributional grammar induction which operate on such principles, using part-of-speech tags as the contextual features. The advantages and disadvantages of these systems are examined, including precision/recall trade-offs, error analysis, and extensibility.
We present a novel generative model for natural language tree structures in which semantic (lexical dependency) and syntactic (PCFG) structures are scored with separate models. This factorization provides conceptual simplicity, straightforward opportunities for separately improving the component models, and a level of performance comparable to similar, non-factored models. Most importantly, unlike other modern parsing models, the factored model admits an extremely effective A* parsing algorithm, which enables efﬁcient, exact inference.
We present an improved method for clustering in the presence of very limited supervisory information, given as pairwise instance constraints. By allowing instance-level constraints to have spacelevel inductive implications, we are able to successfully incorporate constraints for a wide range of data set types. Our method greatly improves on the previously studied constrained -means algorithm, generally requiring less than half as many constraints to achieve a given accuracy on a range of real-world data, while also being more robust when over-constrained. (...) We additionally discuss an active learning algorithm which increases the value of constraints even further. (shrink)
erative clustering. First, we show formally that the common heuristic agglomerative clustering algorithms – Ward’s method, single-link, complete-link, and a variant of group-average – are each equivalent to a hierarchical model-based method. This interpretation gives a theoretical explanation of the empirical behavior of these algorithms, as well as a principled approach to resolving practical issues, such as number of clusters or the choice of method. Second, we show how a model-based viewpoint can suggest variations on these basic agglomerative algorithms. We (...) introduce adjusted complete-link, Mahalanobis-link, and line-link as variants, and demonstrate their utility. (shrink)
This paper presents a novel approach to the unsupervised learning of syntactic analyses of natural language text. Most previous work has focused on maximizing likelihood according to generative PCFG models. In contrast, we employ a simpler probabilistic model over trees based directly on constituent identity and linear context, and use an EM-like iterative procedure to induce structure. This method produces much higher quality analyses, giving the best published results on the ATIS dataset.
While symbolic parsers can be viewed as deduction systems, this view is less natural for probabilistic parsers. We present a view of parsing as directed hypergraph analysis which naturally covers both symbolic and probabilistic parsing. We illustrate the approach by showing how a dynamic extension of Dijkstra’s algorithm can be used to construct a probabilistic chart parser with an Ç´Ò¿µ time bound for arbitrary PCFGs, while preserving as much of the ﬂexibility of symbolic chart parsers as allowed by the inherent (...) ordering of probabilistic dependencies. (shrink)
This paper presents empirical studies and closely corresponding theoretical models of the performance of a chart parser exhaustively parsing the Penn Treebank with the Treebank’s own CFG grammar. We show how performance is dramatically affected by rule representation and tree transformations, but little by top-down vs. bottom-up strategies. We discuss grammatical saturation, including analysis of the strongly connected components of the phrasal nonterminals in the Treebank, and model how, as sentence length increases, the effective grammar rule size increases as regions (...) of the grammar are unlocked, yielding super-cubic observed time behavior in some conﬁgurations. (shrink)
This paper discusses ensembles of simple but heterogeneous classiﬁers for word-sense disambiguation, examining the Stanford-CS224N system entered in the SENSEVAL-2 English lexical sample task. First-order classiﬁers are combined by a second-order classiﬁer, which variously uses majority voting, weighted voting, or a maximum entropy model. While individual ﬁrst-order classiﬁers perform comparably to middle-scoring teams’ systems, the combination achieves high performance. We discuss trade-offs and empirical performance. Finally, we present an analysis of the combination, examining how ensemble performance depends on error independence (...) and task difﬁculty. (shrink)
Robert Rupert is well-known as an vigorous opponent of the hypothesis of extended cognition (HEC). His Cognitive Systems and the Extended Mind is a first-rate development of his “systems-based” approach to demarcating the mind. The results are impressive. Rupert’s account brings much-needed clarity to the often-frustrating debate over HEC: much more than just an attack on HEC, he gives a compelling picture of why the debate matters.
It would be nice if our definition of ‘physical’ incorporated the distinctive content of physics. Attempts at such a definition quickly run into what’s known as Hempel’s dilemma. Briefly: when we talk about ‘physics’, we refer either to current physics or to some idealized version of physics. Current physics is likely wrong and so an unsuitable basis for a definition. ‘Ideal physics’ can’t itself be cashed out except as the science which has completed an accurate survey of the physical; appeals (...) to it to define the physical must therefore end up trivial or circular. So defining the physical in terms of physics looks like a non-starter. (shrink)
Amputation of a limb can result in the persistent hallucination that the limb is still present [Ramachandran and Hirstein, 1998]. Distressingly, these socalled ‘phantom limbs’ are often quite painful. Of a friend whose arm had been amputated due to gas gangrene, W.K. Livingston writes: I once asked him why the sense of tenseness in the hand was so frequently emphasized among his complaints. He asked me to clench my fingers over my thumb, flex my wrist, and raise the arm into (...) a hammerlock position and hold it there. He kept me in this position as long as I could stand it. At the end of five minutes I was perspiring freely, my hand and arm felt unbearably cramped, and I quit. But you can take your hand down, he said. (quoted in [Melzack, 1973] 53) In addition to the obvious medical issues, phantom limb pain also presents philosophical problems. Here’s a thorny one: are phantom limb pains hallucinations of pain? (shrink)
In this paper I discuss philosophical and psychological treatments of the question “how do we decide that an occurrent mental state is a memory and not, say a thought or imagination?” This issue has proven notoriously difficult to resolve, with most proposed indices, criteria and heuristics failing to achieve consensus. Part of the difficulty, I argue, is that the indices and analytic solutions thus far offered seldom have been situated within a well-specified theory of memory function. As I hope to (...) show, when such an approach is adopted, not only do new, functionally-grounded answers emerge; we also gain insight into the adaptive significance of the processes that underwrite our belief in the memorial status of a mental state. (shrink)
Cognitive Task Analysis (CTA) helps researchers understand how cognitive skills and strategies make it possible for people to act effectively and get things done. CTA can yield information people needemployers faced with personnel issues, market researchers who want to understand the thought processes of consumers, trainers and others who design instructional systems, health care professionals who want to apply lessons learned from errors and accidents, systems analysts developing user specifications, and many other professionals. CTA can show what makes the workplace (...) workand what keeps it from working as well as it might.Working Minds is a true handbook, offering a set of tools for doing CTA: methods for collecting data about cognitive processes and events, analyzing them, and communicating them effectively. It covers both the "why" and the "how" of CTA methods, providing examples, guidance, and stories from the authors' own experiences as CTA practitioners. Because effective use of CTA depends on some conceptual grounding in cognitive theory and researchon knowing what a cognitive perspective can offerthe book also offers an overview of current research on cognition.The book provides detailed guidance for planning and carrying out CTA, with chapters on capturing knowledge and capturing the way people reason. It discusses studying cognition in real-world settings and the challenges of rapidly changing technology. And it describes key issues in applying CTA findings in a variety of fields. Working Minds makes the methodology of CTA accessible and the skills involved attainable. (shrink)
In this paper I argue that radiological attempts to elucidate the properties of self -- an endeavor currently popular in the social neurosciences -- are fraught with conceptual difficulties. I first discuss several philosophical criteria that increase the chances we are posing the “right” questions to nature. I then discuss whether these criteria are met when empirical efforts are directed at one of the central constructs in the social sciences – the human self. In particular, I consider whether recent attempts (...) to map the neural correlates of self and its assumed properties using brain scanning technology satisfy the conceptual conditions minimally required to ask well-formed, theoretically satisfying questions of nature. I conclude that much theoretical work remains to be done. (shrink)
Common wisdom, philosophical analysis and psychological research share the view that memory is subjectively positioned toward the past: Specifically, memory enables one to become re-acquainted with the objects and events of his or her past. In this paper I call this assumption into question. As I hope to show, memory has been designed by natural selection not to relive the past, but rather to anticipate and plan for future contingencies -- a decidedly future-oriented mode of subjective temporality. This is not (...) to say memory makes no reference to the past. But, I argue, past-oriented subjectivity is a by-product of a system designed by natural selection to help us face and respond to the “now and the next”. I discuss the implications of the proposed temporal realignment for research agendas as well as the potential limitations of measures designed to explore memory by focusing on its retentive capabilities. (shrink)
The Two Selves takes the position that the self is not a "thing" easily reduced to an object of scientific analysis. Rather, the self consists in a multiplicity of aspects, some of which have a neuro-cognitive basis (and thus are amenable to scientific inquiry) while other aspects are best construed as first-person subjectivity, lacking material instantiation. As a consequence of their potential immateriality, the subjective aspect of self cannot be taken as an object and therefore is not easily amenable to (...) treatment by current scientific methods. -/- Klein argues that to fully appreciate the self, its two aspects must be acknowledged, since it is only in virtue of their interaction that the self of everyday experience becomes a phenomenological reality. However, given their different metaphysical commitments (i.e., material and immaterial aspects of reality), a number of issues must be addressed. These include, but are not limited to, the possibility of interaction between metaphysically distinct aspects of reality, questions of causal closure under the physical, the principle of energy conservation. -/- After addressing these concerns, Klein presents evidence based on self-reports from case studies of individuals who suffer from a chronic or temporary loss of their sense of personal ownership of their mental states. Drawing on this evidence, he argues that personal ownership may be the factor that closes the metaphysical gap between the material and immaterial selves, linking these two disparate aspects of reality, thereby enabling us to experience a unified sense of self despite its underlying multiplicity. -/- . (shrink)
Multiply realizable properties are those whose realizers are physically diverse. It is often argued that theories which contain them are ipso facto irreducible. These arguments assume that physical explanations are restricted to the most specific descriptions possible of physical entities. This assumption is descriptively false, and philosophically unmotivated. I argue that it is a holdover from the late positivist axiomatic view of theories. A semantic view of theories, by contrast, correctly allows scientific explanations to be couched in the most perspicuous, (...) powerful language available. On a semantic view, traditional notions of multiple realizability are thus very hard to motivate. At best, one must abandon either the idea that multiple realizability is an interesting scientific notion, or else admit that multiply realizable properties do not automatically block scientific reductions. (shrink)
In early numerical development, children have to become familiar with the Arabic number system and its place-value structure. The present review summarizes and discusses evidence for language influences on the acquisition of the highly transparent structuring principles of digital-Arabic digits by means of its moderation through the transparency of the respective language's number word system. In particular, the so-called inversion property (i.e., 24 named as "four and twenty" instead of "twenty four") was found to influence number processing in children not (...) only in verbal but also in non-verbal numerical tasks. Additionally, there is first evidence suggesting that inversion-related difficulties may influence numerical processing longitudinally. Generally, language-specific influences in children's numerical development are most pronounced for multi-digit numbers. Yet, there is currently only one study on three-digit number processing for German-speaking children. A direct comparison of additional new data from Italian-speaking children further corroborates the assumption that language impacts on cognitive (number) processing as inversion-related interference was found most pronounced for German-speaking children. In sum, we conclude that numerical development may not be language-specific but seems to be moderated by language. (shrink)
This article explores the connection between the Heng Xian and the Changes of Zhou tradition, especially the “Tuan” and “Attached Verbalizations” commentaries. Two important Heng Xian terms—heng 恆 and fu 復—are also Changes of Zhou hexagrams and possible connections are explored. Second, the Heng Xian account of the creation of names is compared with the “Attached Verbalizations” account of the creation of the Changes of Zhou system. Third, the roles played by knowing and desire in both Heng Xian and the (...) Changes of Zhou tradition are explored, with particular focus on potential points of similarity. Finally, insights gained through these comparisons are used to interpret the Heng Xian advice on initiating action. (shrink)
Infinitism in Epistemology This article provides an overview of infinitism in epistemology. Infinitism is a family of views in epistemology about the structure of knowledge and epistemic justification. It contrasts naturally with coherentism and foundationalism. All three views agree that knowledge or justification requires an appropriately structured chain of reasons. What form may such a […].
Episodic memory often is conceptualized as a uniquely human system of long-term memory that makes available knowledge accompanied by the temporal and spatial context in which that knowledge was acquired. Retrieval from episodic memory entails a form of first–person subjectivity called autonoetic consciousness that provides a sense that a recollection was something that took place in the experiencer’s personal past. In this paper I expand on this definition of episodic memory. Specifically, I suggest that (a) the core features assumed unique (...) to episodic memory are shared by semantic memory, (b) episodic memory cannot be fully understood unless one appreciates that episodic recollection requires the coordinated function of a number of distinct, yet interacting, “enabling” systems. Although these systems – ownership, self, subjective temporality, and agency – are not traditionally viewed as memorial in nature, each is necessary for episodic recollection and jointly they may be sufficient, and (c) the type of subjective awareness provided by episodic recollection (autonoetic) is relational rather than intrinsic – i.e., it can be lost in certain patient populations, thus rendering episodic memory content indistinguishable from the content of semantic long-term memory. (shrink)
Research on future-oriented mental time travel (FMTT) is highly active yet somewhat unruly. I believe this is due, in large part, to the complexity of both the tasks used to test FMTT and the concepts involved. Extraordinary care is a necessity when grappling with such complex and perplexing metaphysical constructs as self and time and their co-instantiation in memory. In this review, I first discuss the relation between future mental time travel and types of memory (episodic and semantic). I then (...) examine the nature of both the types of self-knowledge assumed to be projected into the future and the types of temporalities that constitute projective temporal experience. Finally, I argue that a person lacking episodic memory should nonetheless be able to imagine a personal future by virtue of (a) the fact that semantic, as well as episodic, memory can be self-referential, (b) autonoetic awareness is not a prerequisite for FMTT, and (c) semantic memory does, in fact, enable certain forms of personally-oriented FMTT. (shrink)