A Comparison of ChatGPT and Fine-Tuned Open Pre-Trained Transformers (OPT) Against Widely Used Sentiment Analysis Tools: Sentiment Analysis of COVID-19 Survey Data.

BACKGROUND: Health care providers and health-related researchers face significant challenges when applying sentiment analysis tools to health-related free-text survey data. Most state-of-the-art applications were developed in domains such as social media, and their performance in the health care context remains relatively unknown. Moreover, existing studies indicate that these tools often lack accuracy and produce inconsistent results. OBJECTIVE: This study aims to address the lack of comparative analysis on sentiment analysis tools applied to health-related free-text survey data in the context of COVID-19. The objective was to automatically predict sentence sentiment for 2 independent COVID-19 survey data sets from the National Institutes of Health and Stanford University. METHODS: Gold standard labels were created for a subset of each data set using a panel of human raters. We compared 8 state-of-the-art sentiment analysis tools on both data sets to evaluate variability and disagreement across tools. In addition, few-shot learning was explored by fine-tuning Open Pre-Trained Transformers (OPT; a large language model [LLM] with publicly available weights) using a small annotated subset and zero-shot learning using ChatGPT (an LLM without available weights). RESULTS: The comparison of sentiment analysis tools revealed high variability and disagreement across the evaluated tools when applied to health-related survey data. OPT and ChatGPT demonstrated superior performance, outperforming all other sentiment analysis tools. Moreover, ChatGPT outperformed OPT, exhibited higher accuracy by 6% and higher F-measure by 4% to 7%. CONCLUSIONS: This study demonstrates the effectiveness of LLMs, particularly the few-shot learning and zero-shot learning approaches, in the sentiment analysis of health-related survey data. These results have implications for saving human labor and improving efficiency in sentiment analysis tasks, contributing to advancements in the field of automated sentiment analysis.

Emotion and motion: Toward emotion recognition based on standing and walking.

Emotion recognition is key to interpersonal communication and to human-machine interaction. Body expression may contribute to emotion recognition, but most past studies focused on a few motions, limiting accurate recognition. Moreover, emotions in most previous research were acted out, resulting in non-natural motion, which is unapplicable in reality. We present an approach for emotion recognition based on body motion in naturalistic settings, examining authentic emotions, natural movement, and a broad collection of motion parameters. A lab experiment using 24 participants manipulated participants' emotions using pretested movies into five conditions: happiness, relaxation, fear, sadness, and emotionally-neutral. Emotion was manipulated within subjects, with fillers in between and a counterbalanced order. A motion capture system measured posture and motion during standing and walking; a force plate measured center of pressure location. Traditional statistics revealed nonsignificant effects of emotions on most motion parameters; only 7 of 229 parameters demonstrate significant effects. Most significant effects are in parameters representing postural control during standing, which is consistent with past studies. Yet, the few significant effects suggest that it is impossible to recognize emotions based on a single motion parameter. We therefore developed machine learning models to classify emotions using a collection of parameters, and examined six models: k-nearest neighbors, decision tree, logistic regression, and the support vector machine with radial base function and linear and polynomial functions. The decision tree using 25 parameters provided the highest average accuracy (45.8%), more than twice the random guess for five conditions, which advances past studies demonstrating comparable accuracies, due to our naturalistic setting. This research suggests that machine learning models are valuable for emotion recognition in reality and lays the foundation for further progress in emotion recognition models, informing the development of recognition devices (e.g., depth camera), to be used in home-setting human-machine interactions.

Regulatory fit intensifies moral predispositions.

The experience of regulatory fit shapes subsequent moral conduct by intensifying moral predispositions. Results of seven studies (n = 3,559) show that individuals experiencing regulatory fit versus nonfit are more likely to behave in manners consistent with their moral predispositions as assessed by the trait Moral Disengagement scale, the Machiavellianism scale, and the Honesty-Humility Subscale of the HEXACO-60 inventory. Following an experience of regulatory fit (vs. nonfit), participants with moral predispositions were more likely to consider the decision to engage in sexual intercourse outside a committed relationship as a moral issue (Study 1), and were less willing to do so (Study 2); they also expressed higher intentions of reporting income honestly for tax purposes (Study 3), imposed harsher punishment on a transgressor (Study 4), and self-sacrificed more for the common good in a social dilemma (Studies 4 and 5). The opposite was observed when participants with immoral predispositions experienced regulatory fit (vs. nonfit). In an incentivized sender-receiver game, participants with moral predispositions were less likely to lie for monetary gains when they experienced regulatory fit (vs. nonfit), whereas those with immoral predispositions were more likely to lie (Study 7). By operationalizing the regulatory fit experience as incidental to the moral decision context and assessing moral predispositions with at least a week lead or lag from the main experiment, the findings provide unambiguous evidence that regulatory fit impacts moral conduct by intensifying moral predispositions. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

Age-Related Differences in Experiences With Social Distancing at the Onset of the COVID-19 Pandemic: A Computational and Content Analytic Investigation of Natural Language From a Social Media Survey.

BACKGROUND: As COVID-19 poses different levels of threat to people of different ages, health communication regarding prevention measures such as social distancing and isolation may be strengthened by understanding the unique experiences of various age groups. OBJECTIVE: The aim of this study was to examine how people of different ages (1) experienced the impact of the COVID-19 pandemic and (2) their respective rates and reasons for compliance or noncompliance with social distancing and isolation health guidance. METHODS: We fielded a survey on social media early in the pandemic to examine the emotional impact of COVID-19 and individuals' rates and reasons for noncompliance with public health guidance, using computational and content analytic methods of linguistic analysis. RESULTS: A total of 17,287 participants were surveyed. The majority (n=13,183, 76.3%) were from the United States. Younger (18-31 years), middle-aged (32-44 years and 45-64 years), and older (≥65 years) individuals significantly varied in how they described the impact of COVID-19 on their lives, including their emotional experience, self-focused attention, and topical concerns. Younger individuals were more emotionally negative and self-focused, while middle-aged people were other-focused and concerned with family. The oldest and most at-risk group was most concerned with health-related terms but were lower in anxiety (use of fewer anxiety-related terms) and higher in the use of emotionally positive terms than the other less at-risk age groups. While all groups discussed topics such as acquiring essential supplies, they differentially experienced the impact of school closures and limited social interactions. We also found relatively high rates of noncompliance with COVID-19 prevention measures, such as social distancing and self-isolation, with younger people being more likely to be noncompliant than older people (P<.001). Among the 43.1% (n=7456) of respondents who did not fully comply with health orders, people differed substantially in the reasons they gave for noncompliance. The most common reason for noncompliance was not being able to afford to miss work (n=4273, 57.3%). While work obligations proved challenging for participants across ages, younger people struggled more to find adequate space to self-isolate and manage their mental and physical health; middle-aged people had more concerns regarding childcare; and older people perceived themselves as being able to take sufficient precautions. CONCLUSIONS: Analysis of natural language can provide insight into rapidly developing public health challenges like the COVID-19 pandemic, uncovering individual differences in emotional experiences and health-related behaviors. In this case, our analyses revealed significant differences between different age groups in feelings about and responses to public health orders aimed to mitigate the spread of COVID-19. To improve public compliance with health orders as the pandemic continues, health communication strategies could be made more effective by being tailored to these age-related differences.

Culture and medical decision making: Healthcare consumer perspectives in Japan and the United States.

OBJECTIVE: Two studies identified core value influences on medical decision-making processes across and within cultures. METHODS: In Study 1, Japanese and American adults reported desired levels of medical decision-making influence across conditions that varied in seriousness. Cultural antecedents (interdependence, independence, and power distance) were also measured. In Study 2, American adults reviewed a colorectal cancer screening decision aid. Decision preparedness was measured along with interdependence, independence, and desire for medical information. RESULTS: In Study 1, higher interdependence predicted stronger desire for decision-making information in both countries, but was significantly stronger in Japan. The path from information desire to decision-making influence desire was significant only in Japan. The independence path to desire for decision-making influence was significant only in the United States. Power distance effects negatively predicted desire for decision-making influence only in the United States. For Study 2, high (low) interdependents and women (men) in the United States felt that a colorectal cancer screening decision aid helped prepare them more (less) for a medical consultation. Low interdependent men were at significantly higher risk for low decision preparedness. CONCLUSIONS: Study 1 suggests that Japanese participants may tend to view medical decision-making influence as an interdependent, information sharing exchange, whereas American respondents may be more interested in power sharing that emphasizes greater independence. Study 2 demonstrates the need to assess value influences on medical decision-making processes within and across cultures and suggests that individually tailored versions of decision aids may optimize decision preparedness.

Bringing the frame into focus: the influence of regulatory fit on processing fluency and persuasion.

This research demonstrates that people's goals associated with regulatory focus moderate the effect of message framing on persuasion. The results of 6 experiments show that appeals presented in gain frames are more persuasive when the message is promotion focused, whereas loss-framed appeals are more persuasive when the message is prevention focused. These regulatory focus effects suggesting heightened vigilance against negative outcomes and heightened eagerness toward positive outcomes are replicated when perceived risk is manipulated. Enhanced processing fluency leading to more favorable evaluations in conditions of compatibility appears to underlie these effects. The findings underscore the regulatory fit principle that accounts for the persuasiveness of message framing effects and highlight how processing fluency may contribute to the "feeling right" experience when the strategy of goal pursuit matches one's goal.

Primed interference: the cognitive and behavioral costs of an incongruity between chronic and primed motivational orientations.

Research has shown that temporarily primed motivational orientations have essentially the same effects on how people pursue their goals as their chronic orientations. This article shows that, despite the interchangeability of temporary and chronic motivations, primed motivational orientations that are incongruent with chronic orientations create interference, requiring the deployment of cognitive resources and thus undermining performance on subsequent tasks that rely on these resources. Across 6 studies, we primed motivational orientations that were either congruent or incongruent with participants' chronic orientations and then assessed their performance on subsequent tasks that required cognitive resources. Consistent with the primed interference hypothesis, we found that incongruity between temporary and chronic motivational orientations undermined participants' (a) inhibition of incorrect but highly accessible responses, (b) mental arithmetic, (c) analytical reasoning, and (d) resistance to temptation. These results--which were observed following the activation of motivations for promotion or prevention (Studies 1-2 and 5-6), high or low need for belonging (Study 3), and high or low power orientations (Study 4)--illustrate the broad implications of holding incongruent chronic and primed orientations.

When a threat to the brand is a threat to the self: the importance of brand identification and implicit self-esteem in predicting defensiveness.

This research examines how people respond when a commercial brand they identify with is threatened. Across four studies, the authors found that among participants who identified with a brand, a threat to the brand elicited the same responses as a threat to the self. Specifically, participants with low implicit self-esteem defended the brand when the self was activated, unlike their high implicit self-esteem counterparts. In addition, brand defense was reduced when individuals had the opportunity to affirm a valued aspect of their self-concept. These findings suggest that when a brand that people identify with is threatened, they may defend the brand to preserve the integrity of the self. More broadly, these findings are consistent with the notion that brands may be included into the extended self-concept, which supports William James's original ideas concerning the breadth and heterogeneity of the self.