Group-level Factor Research Articles

Evidence of interrelated cognitive-like capabilities in large language models: Indications of artificial general intelligence or achievement?

Large language models (LLMs) are advanced artificial intelligence (AI) systems that can perform a variety of tasks commonly found in human intelligence tests, such as defining words, performing calculations, and engaging in verbal reasoning. There are also substantial individual differences in LLM capacities. Given the consistent observation of a positive manifold and general intelligence factor in human samples, along with group-level factors (e.g., crystallised intelligence), we hypothesized that LLM test scores may also exhibit positive inter-correlations, which could potentially give rise to an artificial general ability (AGA) factor and one or more group-level factors. Based on a sample of 591 LLMs and scores from 12 tests aligned with fluid reasoning (Gf), domain-specific knowledge (Gkn), reading/writing (Grw), and quantitative knowledge (Gq), we found strong empirical evidence for a positive manifold and a general factor of ability. Additionally, we identified a combined Gkn/Grw group-level factor. Finally, the number of LLM parameters correlated positively with both general factor of ability and Gkn/Grw factor scores, although the effects showed diminishing returns. We interpreted our results to suggest that LLMs, like human cognitive abilities, may share a common underlying efficiency in processing information and solving problems, though whether LLMs manifest primarily achievement/expertise rather than intelligence remains to be determined. Finally, while models with greater numbers of parameters exhibit greater general cognitive-like abilities, akin to the connection between greater neuronal density and human general intelligence, other characteristics must also be involved.

Using group level factor models to resolve high dimensionality in model-based sampling.

Joint modeling of decisions and neural activation poses the potential to provide significant advances in linking brain and behavior. However, methods of joint modeling have been limited by difficulties in estimation, often due to high dimensionality and simultaneous estimation challenges. In the current article, we propose a method of model estimation that draws on state-of-the-art Bayesian hierarchical modeling techniques and uses factor analysis as a means of dimensionality reduction and inference at the group level. This hierarchical factor approach can adopt any model for the individual and distill the relationships of its parameters across individuals through a factor structure. We demonstrate the significant dimensionality reduction gained by factor analysis and good parameter recovery, and illustrate a variety of factor loading constraints that can be used for different purposes and research questions, as well as three applications of the method to previously analyzed data. We conclude that this method provides a flexible and usable approach with interpretable outcomes that are primarily data-driven, in contrast to the largely hypothesis-driven methods often used in joint modeling. Although we focus on joint modeling methods, this model-based estimation approach could be used for any high dimensional modeling problem. We provide open-source code and accompanying tutorial documentation to make the method accessible to any researchers. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Attendance, Weight Loss, and Participation in a Behavioural Diabetes Prevention Programme

BackgroundWeight loss in diabetes prevention programmes has been shown to be associated with participants’ age, socio-economic status, and ethnicity. However, little is known about how these differences relate to attendance and whether such differences can be mediated by other potentially modifiable factors. Differential effectiveness across these factors may exacerbate health inequalities.MethodProspective analysis of participant data collected by one provider of the standardised national NHS diabetes prevention programme in England. Mediation analysis was performed via a structural equation model to examine whether the number of attended sessions mediated the associations of age, socio-economic status, and ethnicity with follow-up weight. The group-level factor of number of attended sessions was examined using multiple linear regression as a benchmark; multilevel linear regression using three levels (venue, coach, and group); and fixed effects regression to account for venue-specific and coach-specific characteristics.ResultsThe associations of age, socio-economic status, and ethnicity with follow-up weight were all mediated by the number of attended sessions. Group size was associated with attendance in an inverted ‘U’ shape, and the number of days between referral and group start was negatively associated with attendance. Time of day, day of the week, and the number of past groups led by the coach were not associated with attendance.ConclusionMost of the differences in weight loss initially attributed to socio-demographic factors are mediated by the attendance of the diabetes prevention programme. Therefore, targeted efforts to improve uptake and adherence to such programmes may help alleviate inequalities.

Neighborhood representative for improving outlier detectors

Over the decades, traditional outlier detectors have ignored the group-level factor when calculating outlier scores for objects in data by evaluating only the object-level factor, failing to capture the collective outliers. To mitigate this issue, we present a framework called neighborhood representative (NR), which empowers all the existing outlier detectors to efficiently detect outliers, including collective outliers, while maintaining their computational integrity. It achieves this by selecting representative objects, scoring these objects, then applies the score of the representative objects to its collective objects. Without altering existing detectors, NR is compatible with existing detectors, while improving performance on eleven real world datasets with +8 % (0.72 to 0.78 AUC) on average relative to twelve state-of-the-art outlier detectors. The implementation of NR can be found via www.OutlierNet.com for reproducibility.

호텔에서의 안전리더십과 직원의 안전행동

Safety behavior is highly likely to be influenced by not only individual level factors, but also organizational or group level factors. However, in Korea, few studies have been conducted to classify each of factors on safety behavior. This study examines the effect of team leader’s safety leadership as a group level factor on safety behavior mediated through safety motivation and team member’s safety attitude as an individual level factor. This study was conducted in China from January to June 2022 with 512 subjects from 95 teams of 5-star hotels. Hierarchical Linear Modeling (HLM) for multi-level analysis is used to analyze the mechanisms by which safety leadership and safety attitudes influence safety behavior through safety motivation. The research results indicate that safety leadership and safety attitude can predict safety behavior. Safety motivation also played a mediating role in the relationship between safety leadership and safety behavior. This study, which conducted the time-lagged study, suggests that management should build appropriate safety leadership, continuously check the effectiveness of leadership through regular monitoring, and improve safety-related performance through encouraging safety attitudes and motivation.

Understanding peer pressure on joint consumption decisions: the role of social capital during emerging adulthood

PurposeThe purpose of this paper is to examine the effect of peer pressure on joint consumption decisions among emerging adults. Building on prospect theory and characteristics of emerging adulthood, the authors propose that influence from peers (i.e. informational and normative influence) serves as a channel to understand how peer pressure shapes joint consumer behaviors at different levels of social capital.Design/methodology/approachAn online survey is distributed to the emerging adults, aged 18 to 25, in the south, west, east and middle of the USA. Construct validity and reliability are tested by using confirmatory factor analysis. Structural equation modeling is used to test the mediating and moderating effects.FindingsThe results show that social capital moderates the relationship between peer pressure and group-oriented consumer decisions, such that the relationship is positive in groups with high-level social capital but negative in groups with low-level social capital. Furthermore, such effects tend to be achieved via peer influence. And peer influence is stronger in groups with high-level social capital than those with low-level social capital.Originality/valueThe current literature has shown contradictory results: it is usually believed that emerging adults may conform to pressure and engage in group-oriented decisions; however, some research has reported the opposite result. To better understand this relationship, the authors aim at a group-level factor – perceived social capital – as a boundary condition. This research contributes to the young consumer decision-making literature by involving the interplay among peer pressure, perceived social capital and peer informational and normative influence.

At the core of cyberaggression: A group-based explanation.

In this study, prototypicality of the aggressor was tested as a group‐level factor predicting social media users’ active participation in cyberaggression. Participants were exposed to a fictitious conversation, in which either a prototypical versus non‐prototypical user posted an aggressive comment as a reply to a provocative comment. In line with self‐categorization theory, we hypothesized that bystander participants would post an aggressive comment and rate the aggression as acceptable to a greater extent in the prototypical than in the non‐prototypical condition. Furthermore, we predicted that perceived normativity of aggression would mediate the effect of prototypicality. Results supported these predictions and showed that prototypical members affect the extent to which collective aggressive behaviors in online interactions are approved and enacted. These findings highlight the importance of group‐level factors in the study of cyberaggression and provide important information for understanding the psychological underpinnings of collective forms of online aggression.

Simultaneous training on overlapping grapheme phoneme correspondences augments learning and retention

An important component of learning to read is the acquisition of letter-to-sound mappings. The sheer quantity of mappings and many exceptions in opaque languages such as English suggests that children may use a form of statistical learning to acquire them. However, whereas statistical models of reading are item-based, reading instruction typically focuses on rule-based approaches involving small sets of regularities. This discrepancy poses the question of how different groupings of regularities, an unexamined factor of most reading curricula, may affect learning. Exploring the interplay between item statistics and rules, this study investigated how consonant variability, an item-level factor, and the degree of overlap among the to-be-trained vowel strings, a group-level factor, influence learning. English-speaking first graders (N = 361) were randomly assigned to be trained on vowel sets with high overlap (e.g., EA, AI) or low overlap (e.g., EE, AI); this was crossed with a manipulation of consonant frame variability. Whereas high vowel overlap led to poorer initial performance, it resulted in more learning when tested immediately and after a 2-week-delay. There was little beneficial effect of consonant variability. These findings indicate that online letter/sound processing affects how new knowledge is integrated into existing information. Moreover, they suggest that vowel overlap should be considered when designing reading curricula.

Neural Representations of Abstract Concepts: Identifying Underlying Neurosemantic Dimensions

The abstractness of concepts is sometimes defined indirectly as lacking concreteness, this view provides little insight into their cognitive or neural basis. Multivariate pattern analytic techniques applied to functional magnetic resonance imaging data were used to characterize the neural representations of 28 individual abstract concepts. A classifier trained on the concepts' neural signatures reliably decoded their neural representations in an independent subset of data for each participant. There was considerable commonality of the neural representations across participants as indicated by the accurate classification of each participant's concepts based on the neural signatures obtained in other participants. Group-level factor analysis revealed 3 semantic dimensions underlying the 28 concepts, suggesting a brain-based ontology for this set of abstract concepts. The 3 dimensions corresponded to 1) the degree a concept was Verbally Represented; 2) whether a concept was External (or Internal) to the individual, and 3) whether the concept contained Social Content. Further exploration of the Verbal Representation dimension suggests that the degree a concept is verbally represented can be construed as a point on a continuum between language faculties and perceptual faculties. A predictive model, based on independent behavioral ratings of the 28 concepts along the 3 factor dimensions, provided converging evidence for the interpretations.

On the Unethical Use of Privileged Information in Strategic Decision-Making: The Effects of Peers’ Ethicality, Perceived Cohesion, and Team Performance

In order to make strategic decisions and improve their firm’s performance, top management teams must have information on the competitive context in general, and the firm’s competitors in particular. During the decision-making process, top managers can have access to “privileged information”—i.e., information of a confidential and potentially strategic nature that could ultimately confer a decisional advantage over competing parties. However, obtaining and using privileged information in a business context is often illegal—and if not, is usually deemed unethical or “against the rules.” Using a quasi-experimental design, this study explores the reasons why an individual might engage in such unethical behavior. We assess the extent to which managers use privileged information with respect to perceived team cohesion and peers’ ethicality. More specifically, our results show that the use of privileged information is predicted by the decision-maker’s perceptions of their team cohesion and their peers’ ethicality. Moreover, we find that team performance, as a group-level nonself-reported factor (measured by the firm’s share price in our simulation), moderates the relationship between cohesion and the use of privileged information. The relationship between cohesion, ethical behavior, and team performance is also discussed. We draw on these findings to make some practical suggestions on how to incorporate practices that could better prevent the unethical use of privileged information in strategic decision-making processes.

Comment on “Residual group-level factor associations: Possibly negative implications for the mutualism theory of general intelligence” by Gilles E. Gignac (2016)

We show by simulations that the main premise underlying Gignac's (2016) rejection of the mutualism model of general intelligence is incorrect. We discuss our results in the context of the debate on g-factor theory and its alternatives, especially with regard to the claim that g-factor theory is much simpler than mutualism theory.

Residual group-level factor associations: Possibly negative implications for the mutualism theory of general intelligence

The mutualism theory of general intelligence (g) posits that the positive manifold arises because of mutually beneficial interactions between originally orthogonal cognitive abilities, rather than because of a genuine general construct. In a recent investigation, Gignac (2014) reported that the strength of g was largely constant from the ages of 2.5 to 90years, which was interpreted as an indirect failure to confirm the mutualism theory of g. In this investigation, a second indirect test of the mutualism theory of g was performed. Specifically, it was hypothesized that, if the extended mutualism theory of g is plausible, then there should be some consistent, positive associations between group-level factors, controlling for the effects of a general factor. To examine this possibility, the associations between cognitive ability group-level factors were estimated across a series of seven relatively large and relatively representative samples of intelligence battery data. Next, seven single-factor models were estimated against the seven group-level inter-correlation matrices. The corresponding single-factor residual correlation matrices were observed to yield an approximately equal number of positive and negative residual correlations and an overall mean of zero. Furthermore, the only moderately consistent residual effect was a negative association between crystallised intelligence and processing speed. Although only an indirect test, the results are interpreted to be more supportive of g factor theory than mutualism.

Raven's is not a pure measure of general intelligence: Implications for g factor theory and the brief measurement of g

It has been claimed that Raven's Progressive Matrices is a pure indicator of general intelligence (g). Such a claim implies three observations: (1) Raven's has a remarkably high association with g; (2) Raven's does not share variance with a group-level factor; and (3) Raven's is associated with virtually no test specificity. The existing factor analytic research relevant to Raven's and g is very mixed, likely because of the variety of factor analytic techniques employed, as well as the small sample sizes upon which the analyses have been performed. Consequently, the purpose of this investigation was to estimate the association between Raven's and g, Raven's and a theoretically congruent group-level factor, and Raven's test specificity within the context of a bifactor model. Across several large samples, it was observed that Raven's (1) shared approximately 50% of its variance with g; (2) shared approximately 10% of its variance with a fluid intelligence group-level factor orthogonal to g; and (3) was associated with approximately 25% test specific reliable variance. Overall, the results are interpreted to suggest that Raven's is not a particularly remarkable test with respect to g. Potential implications relevant to the commonly articulated central role of Raven's in g factor theory, as well as the Flynn effect, are discussed. Finally, researchers are discouraged to include only Raven's in an investigation, if a valid estimate of g is sought. Instead, as just one example, a four-subtest combination from the Wechsler scales with a g validity coefficient of .93 and 14min administration time is suggested.

Inference of Cross-Level Interaction between Genes and Contextual Factors in a Matched Case-Control Metabolic Syndrome Study: A Bayesian Approach

Genes, environment, and the interaction between them are each known to play an important role in the risk for developing complex diseases such as metabolic syndrome. For environmental factors, most studies focused on the measurements observed at the individual level, and therefore can only consider the gene-environment interaction at the same individual scale. Indeed the group-level (called contextual) environmental variables, such as community factors and the degree of local area development, may modify the genetic effect as well. To examine such cross-level interaction between genes and contextual factors, a flexible statistical model quantifying the variability of the genetic effects across different categories of the contextual variable is in need. With a Bayesian generalized linear mixed-effects model with an unconditional likelihood, we investigate whether the individual genetic effect is modified by the group-level residential environment factor in a matched case-control metabolic syndrome study. Such cross-level interaction is evaluated by examining the heterogeneity in allelic effects under various contextual categories, based on posterior samples from Markov chain Monte Carlo methods. The Bayesian analysis indicates that the effect of rs1801282 on metabolic syndrome development is modified by the contextual environmental factor. That is, even among individuals with the same genetic component of PPARG_Pro12Ala, living in a residential area with low availability of exercise facilities may result in higher risk. The modification of the group-level environment factors on the individual genetic attributes can be essential, and this Bayesian model is able to provide a quantitative assessment for such cross-level interaction. The Bayesian inference based on the full likelihood is flexible with any phenotype, and easy to implement computationally. This model has a wide applicability and may help unravel the complexity in development of complex diseases.

Person-Fit to the Five Factor Model of Personality

The Five Factor Model (FFM), a valid model of interindividual differences in the personality of a group of people, reportedly does not always provide a good fit for the individuals of that group. In addition to intraindividual variation across a considerable period of time, meaningful intraindividual variation can be observed within a single test administration. Two person-fit indices showed that the FFM is an adequate model for 95% of the 1,765 target-judge pairs in four different countries (Belgium, the Czech Republic, Estonia, and Germany): the double-entry intraclass correlation (ICCDE), which indicated that the 30 NEO PI-R scores on scales measuring the same personality trait are more similar and certainly less different than scores measuring different traits, and the individual contribution to the extracted eigenvalues (Zeig). The individual response pattern to the personality questionnaire characterized by the ICCDE and Zeig strongly determined the percentage of explained variance for the group-level factor structure of interindividual differences and the mean self-observer profile agreement. We demonstrate that, if the percentage of variance explained by the first five principal components is high enough, the FFM also provides an adequate fit at the individual level for most people.

Improving Multilevel Analyses

Multilevel analysis has been widely used to allow the simultaneous examination of the effects of individual- and group-level variables on individual health outcomes. In spite of its utility, multilevel design can have some drawbacks in the estimation of risk factor effects when the within-group variation of variables of interest is small relative to between-group variation. An extreme case of this is a group-level risk factor, which by definition has no within-group variation. To improve the estimation of group-level and individual-level risk factor effects, we consider an integrated epidemiologic design using a population-based estimating equation approach that can be considered a further extension of the multilevel design. Although the integrated design uses the same individual-level and group-level data as the multilevel design, it includes aggregated health outcome data in each group as additional information. This paper explains differences between the 2 designs, describing advantages and disadvantages of the integrated design over the multilevel design. The 2 designs are applied to a real example of mortality following chronic kidney disease, illustrating differences that might be encountered in practice.

Civic entrepreneurship in Australia

PurposeThe purpose of this study is to expand on previous research conducted by Hornsby et al. that examined the corporate entrepreneurship internal factor of managerial attitude.Design/methodology/approachThis paper develops and tests a group level factor of knowledge, the explication of tacit knowledge and a factor of managerial attitude, namely leadership support that is inherently multi‐level in nature.FindingsLeadership support is significant at both the dyad level and at the group level of analysis. Ordinary least squares regression supported the main hypothesis, that leadership support has a direct positive impact on explication of tacit knowledge.Research limitations/implicationsSuccessfully testing the group level significance of leadership support has implications for future research because it is considered an individual level variable. Developing and testing the explication of tacit knowledge construct contributes to research on knowing in organisations because it provides a metric that is an indicator of the explication of tacit knowledge.Practical implicationsThese research findings have management implications for the way local government creates innovative top management teams to facilitate local economic and community development.Originality/valueThis paper represents an early contribution to the literature.