Answer comparison in automated question answering

In the context of Question Answering (QA) on free text, we assess the value of answer comparison and information fusion in handling multiple answers. We report improvements in answer re-ranking using fusion on a set of location questions and show the advantages of considering candidates as allies rather than competitors. We conclude with some observations about answer modeling and evaluation methodology, arising from a more recent experiment with a larger set of questions and a greater diversity of question types and candidates.     

Why neuroscience matters to cognitive neuropsychology

The broad issue in this paper is the relationship between cognitive psychology and neuroscience. That issue arises particularly sharply for cognitive neurospsychology, some of whose practitioners claim a methodological autonomy for their discipline. They hold that behavioural data from neuropsychological impairments are sufficient to justify assumptions about the underlying modular structure of human cognitive architecture, as well as to make inferences about its various components. But this claim to methodological autonomy can be challenged on both philosophical and empirical grounds. A priori considerations about (cognitive) multiple realisability challenge the thesis on philosophical grounds, and neuroscientific findings from developmental disorders substantiate that challenge empirically. The conclusion is that behavioural evidence alone is inadequate for scientific progress since appearances of modularity can be thoroughly deceptive, obscuring both the dynamic processes of neural development and the endstate network architecture of real cognitive systems.     

Reliability and validity of the Multiple Affect Adjective Check List - Revised

A recent factor analysis of the Multiple Affect Adjective Check List (MAACL) changed the Anxiety, Depression, and Hostility scales from bipolar to unipolar scales and added two new scales: Positive Affect and Sensation Seeking. Internal reliability of the MAACL-R scales, computed for normal and patient samples, was adequate for state and trait forms except for Sensation Seeking. Test-retest reliability in college students was higher for the trait form with retest intervals of from 2 to 8 weeks than for the state form with retest intervals of from 2 to 5 days. The pattern of correlations among self-, peer, and counselor ratings and the MAACL-R scales for normal and patient samples indicates improved discriminant validity and equally good convergent validity as the old scales. The use of standard scores that are indexed to the number of items checked reduced scale intercorrelations by controlling the acquiescence set.The financial support of the Weldon Springs Endowment Fund, University of Missouri at Kansas City, toward the completion of this research project is gratefully acknowledged.     

Dimensionality of the Multiple Affect Adjective Check List-Revised: A comparison of factor analytic procedures

The purpose of the study was to determine the factor structure of the revised Multiple Affect Adjective Check List (MAACL-R). The data from 307 undergraduate students who completed the MAACL-R were factor analyzed using both principal-components and principal-axis techniques. Based upon research with the original MAACL, two- and five-factor solutions to the MAACL-R factor structure were examined. The principal-components and principal-axis techniques yielded similar item loadings for each solution and the results of both techniques indicated the superiority of the two-factor solution. The two identified factors could be best characterized as representing independent dimensions of negative and positive affect. These findings are interpreted as providing support for the use of the MAACL-R summary scores of Dysphoria and PASS (Positive Affect and Sensation Seeking) rather than for the use of the five MAACL-R subscales.     

Prioritizing—The task strategy of the powerful?

Previous research has shown that power increases focus on the main goal when distractor information is present. As a result, high-power people have been described as goal-focused. In real life, one typically wants to pursue multiple goals at the same time. There is a lack of research on how power affects how people deal with situations in which multiple important goals are present. To address this question, 158 participants were primed with high or low power or assigned to a control condition, and were asked to perform a dual-goal task with three difficulty levels. We hypothesized and found that high-power primed people prioritize when confronted with a multiple-goal situation. More specifically, when task demands were relatively low, power had no effect; participants generally pursued multiple goals in parallel. However, when task demands were high, the participants in the high-power condition focused on a single goal whereas participants in the low-power condition continued using a dual-task strategy. This study extends existing power theories and research in the domain of goal pursuit.     

The role of stimulus ambiguity and movement in spatial navigation: a multiple memory systems analysis of location discrimination

This paper reviews recent findings about how rats navigate by learning to discriminate among locations. The assumption underlying the experiments and their interpretation is that the information required to do this is learned by three independent, parallel memory systems. One system processes cognitive information (or "knowledge"), a second system processes reinforced stimulus-response associations and a third processes Pavlovian conditioned responses in the form of stimulus-affect associations. The information stored in each system produces behavior that, in some cases, results in a location discrimination. The present experiments focus on three factors that influence what each system learns and whether the resulting memory produces behavior that results in a location discrimination. One factor is whether the locations to be discriminated can be identified by unique, unambiguous stimuli or whether they are ambiguously associated with the same stimuli. The second factor is whether the stimuli are observed passively or whether the rats move among them, voluntarily or involuntarily. The third factor is whether or not the rats perform specific reinforced responses in the presence of the stimuli. Instances of co-operative behavioral outputs from memory systems that facilitate location discriminations and of competitive outputs that impede discriminations are described.     

Methodological issues in judgment and decision-making research: Concurrent verbal protocol validity and simultaneous traces of process

This paper examines two dimensions of concurrent verbal protocol validity. First, whether verbalization affects process and outcome (reactivity) was examined by comparing concurrent verbal protocol traces with those from a computer search process tracing method, the latter being a complete trace of information acquisition from experimental materials. Earlier findings that verbalization affects time were confirmed. However, verbalization did not affect amount and pattern of acquisition or accuracy of judgments. Second, whether concurrent verbal protocols are complete was examined by comparing concurrent verbal protocol and computer traces that were simultaneously obtained in a treatment in which subjects verbalized as they acquired information from the computer. The verbal traces less completely captured information acquisition behavior than computer search. This suggests that, although concurrent verbal protocols provide greater insight into decision behavior than computer search, the latter is a more reliable information-acquisition trace. Thus, if information acquisition is of primary interest and if computer search activities can be naturally integrated into performing the primary task, computer search is preferred to concurrent verbal protocols. However, if information use or retrieval from long-term memory is of primary interest, concurrent verbal protocols are preferred to computer search. Finally, this paper examined whether the simultaneous use of concurrent verbal protocols and computer search provides traces of information acquisition and use that are as complete as when each method is independently applied. Although computer search tended to limit subjects verbalizations of evaluative operators, this effect may be eliminated by practice on the computer prior to collecting data.     

EM and beyond

The basic theme of the EM algorithm, to repeatedly use complete-data methods to solve incomplete data problems, is also a theme of several more recent statistical techniques. These techniques—multiple imputation, data augmentation, stochastic relaxation, and sampling importance resampling—combine simulation techniques with complete-data methods to attack problems that are difficult or impossible for EM.A preliminary version of this article was the Keynote Address at the 1987 European Meeting of the Psychometric Society June 24–26, 1987 in Enschede, The Netherlands. The author wishes to thank the editor and reviewers for helpful comments.     

Bayesian Extended Redundancy Analysis: A Bayesian Approach to Component-based Regression with Dimension Reduction

Abstract

Extended redundancy analysis (ERA) combines linear regression with dimension reduction to explore the directional relationships between multiple sets of predictors and outcome variables in a parsimonious manner. It aims to extract a component from each set of predictors in such a way that it accounts for the maximum variance of outcome variables. In this article, we extend ERA into the Bayesian framework, called Bayesian ERA (BERA). The advantages of BERA are threefold. First, BERA enables to make statistical inferences based on samples drawn from the joint posterior distribution of parameters obtained from a Markov chain Monte Carlo algorithm. As such, it does not necessitate any resampling method, which is on the other hand required for (frequentist’s) ordinary ERA to test the statistical significance of parameter estimates. Second, it formally incorporates relevant information obtained from previous research into analyses by specifying informative power prior distributions. Third, BERA handles missing data by implementing multiple imputation using a Markov Chain Monte Carlo algorithm, avoiding the potential bias of parameter estimates due to missing data. We assess the performance of BERA through simulation studies and apply BERA to real data regarding academic achievement.     

Five Phases of Dyadic Analysis: Stretching the Boundaries of Understanding of Family Relationships

In the last decade, an increasing number of qualitative studies sought to investigate the dynamics of various dyads by conducting in-depth, multiple family member interviews. The emphasis in the methodological literature dealing with this type of research is primarily on the data collection process, and much less on the development of methods suitable for the analysis of the data thus derived, especially with regard to dyads consisting of family members belonging to different systems: families of origin or nuclear families. The purpose of this paper is to propose a model for dyadic analysis based on examining the dynamics of the evolving relationships between key caregivers of a family member with brain injury. The model includes inductive and abductive phases of analysis, and it is based on an ecological-systemic perspective. The benefits of this model are highlighted, and its potential contribution is further discussed.