Character-Principlism and the Particularity Objection

This paper is a response to particularist critics of the normative force of moral principles. The particularist critique, as I understand it, is a rejection not only of principle-based accounts of moral deliberation and justification, but also of accounts of character in which principles play a central role. I focus on the latter challenge and counter it with a view I call character-principlism .
I begin by discussing in a general way what motivates the particularity objection to principles and then contrast two views – both of which insist on the importance of attentiveness to particularity – about the relative normative status of principles and particular cases. I present some reasons for believing that we need a more normatively robust conception of the role of moral principles than the particularists provide. In the main portion of the paper, I discuss how character-principlism sees principles functioning in our lives and the lives we lead with others. I contrast this with some other accounts of desirable character that particularists can embrace, and argue that these are seriously flawed because, unlike character-principlism, they cannot satisfactorily explain how a person could possess the constancy of character that moral integrity requires.     

Single-Stage Prediction Models Do Not Explain the Magnitude of Syntactic Disambiguation Difficulty

The disambiguation of a syntactically ambiguous sentence in favor of a less preferred parse can lead to slower reading at the disambiguation point. This phenomenon, referred to as a garden-path effect, has motivated models in which readers initially maintain only a subset of the possible parses of the sentence, and subsequently require time-consuming reanalysis to reconstruct a discarded parse. A more recent proposal argues that the garden-path effect can be reduced to surprisal arising in a fully parallel parser: words consistent with the initially dispreferred but ultimately correct parse are simply less predictable than those consistent with the incorrect parse. Since predictability has pervasive effects in reading far beyond garden-path sentences, this account, which dispenses with reanalysis mechanisms, is more parsimonious. Crucially, it predicts a linear effect of surprisal: the garden-path effect is expected to be proportional to the difference in word surprisal between the ultimately correct and ultimately incorrect interpretations. To test this prediction, we used recurrent neural network language models to estimate word-by-word surprisal for three temporarily ambiguous constructions. We then estimated the slowdown attributed to each bit of surprisal from human self-paced reading times, and used that quantity to predict syntactic disambiguation difficulty. Surprisal successfully predicted the existence of garden-path effects, but drastically underpredicted their magnitude, and failed to predict their relative severity across constructions. We conclude that a full explanation of syntactic disambiguation difficulty may require recovery mechanisms beyond predictability.     

The Role of Emotional Content and Perceptual Saliency During the Programming of Saccades Toward Faces

Previous studies have shown that the human visual system can detect a face and elicit a saccadic eye movement toward it very efficiently compared to other categories of visual stimuli. In the first experiment, we tested the influence of facial expressions on fast face detection using a saccadic choice task. Face-vehicle pairs were simultaneously presented and participants were asked to saccade toward the target (the face or the vehicle). We observed that saccades toward faces were initiated faster, and more often in the correct direction, than saccades toward vehicles, regardless of the facial expressions (happy, fearful, or neutral). We also observed that saccade endpoints on face images were lower when the face was happy and higher when it was neutral. In the second experiment, we explicitly tested the detection of facial expressions. We used a saccadic choice task with emotional-neutral pairs of faces and participants were asked to saccade toward the emotional (happy or fearful) or the neutral face. Participants were faster when they were asked to saccade toward the emotional face. They also made fewer errors, especially when the emotional face was happy. Using computational modeling, we showed that this happy face advantage can, at least partly, be explained by perceptual factors. Also, saccade endpoints were lower when the target was happy than when it was fearful. Overall, we suggest that there is no automatic prioritization of emotional faces, at least for saccades with short latencies, but that salient local face features can automatically attract attention.     

Refuting the unfolding-argument on the irrelevance of causal structure to consciousness

The unfolding argument (UA) was advanced as a refutation of prominent theories, which posit that phenomenal experience is determined by patterns of neural activation in a recurrent (neural) network (RN) structure. The argument is based on the statement that any input–output function of an RN can be approximated by an “equivalent” feedforward-network (FFN). According to UA, if consciousness depends on causal structure, its presence is unfalsifiable (thus non-scientific), as an equivalent FFN structure is behaviorally indistinguishable with regards to any behavioral test. Here I refute UA by appealing to computational theory and cognitive-neuroscience. I argue that a robust functional equivalence between FFN and RN is not supported by the mathematical work on the Universal Approximator theorem, and is also unlikely to hold, as a conjecture, given data in cognitive neuroscience; I argue that an equivalence of RN and FFN can only apply to static functions between input/output layers and not to the temporal patterns or to the network’s reactions to structural perturbations. Finally, I review data indicating that consciousness has functional characteristics, such as a flexible control of behavior, and that cognitive/brain dynamics reveal interacting top-down and bottom-up processes, which are necessary for the mediation of such control processes.     

Learning and development in neural networks--the importance of prior experience

Infants can discriminate between familiar and unfamiliar grammatical patterns expressed in a vocabulary that is distinct from that used earlier during familiarization (Cognition 70(2) (1999) 109; Science 283 (1999) 77). Various models have captured the data, although each required that discrimination be distinct, in terms of the computational process, from familiarization. This article describes a simple recurrent network (SRN), equipped only with the assumption that it should predict what comes next, which models the data without distinguishing between familiarization and discrimination. To accomplish this, the SRN requires pre-training on a range of sequences instantiating different structures and different vocabulary items to those used subsequently during familiarization and test. Pre-training enables the network to avoid replacing structure acquired during familiarization with structure experienced at test. An equivalent enabling condition may underpin infants' resistance to catastrophic interference between the different structures and vocabulary items to which they are exposed.     

Transitive choices by a simple, fully connected, backpropagation neural network: implications for the comparative study of transitive inference.

In search of the minimal requirements for transitive reasoning, a simple neural network was trained and tested on the non-verbal version of the conventional "five-term-series task" – a paradigm used with human adults, children and a variety of non-human species. The transitive performance of the network was analogous in several aspects to that reported for children and animals. The three effects usually associated with transitive choices i.e. "symbolic distance", "lexical marking" and "end-anchor", were also clearly shown by the neural network. In a second experiment, where the training conditions were manipulated, the network failed to match the behavioural pattern reported for human adults in the test following an ordered presentation of the premises. However, it mimicked young children's performance when tested with a novel comparison term. Although we do not intend to suggest a new model of transitive inference, we conclude, in line with other authors, that a simple error-correcting rule can generate transitive behaviour similar to the choice pattern of children and animals in the binary form of the five-term-series task without requiring high-order logical or paralogical abilities. The analysis of the training history and of the final internal structure of the network reveals the associative strategy employed. However, our results indicate that the scope of the associative strategy used by the network might be limited. The extent to which the conventional five-term-series task, in absence of appropriate manipulations of training and testing conditions, is suitable to detect cognitive differences across species is also discussed on the basis of our results. Accepted after revision: 29 May 2001 Electronic Publication     

Why there might be a moral faculty: A reply to Johnson

Is there a cognitive faculty dedicated to the moral domain? Mark Johnson has developed a number of arguments against the existence of such a faculty. I claim that these arguments are not persuasive and that there may be a moral faculty.     

The natural mathematics of behavior analysis

Models that generate event records have very general scope regarding the dimensions of the target behavior that we measure. From a set of predicted event records, we can generate predictions for any dependent variable that we could compute from the event records of our subjects. In this sense, models that generate event records permit us a freely multivariate analysis. To explore this proposition, we conducted a multivariate examination of Catania's Operant Reserve on single VI schedules in transition using a Markov Chain Monte Carlo scheme for Approximate Bayesian Computation. Although we found systematic deviations between our implementation of Catania's Operant Reserve and our observed data (e.g., mismatches in the shape of the interresponse time distributions), the general approach that we have demonstrated represents an avenue for modelling behavior that transcends the typical constraints of algebraic models.