In a recent article published in PNAS, researchers argued that an individual’s behavior and socioeconomic circumstances are associated with their well-being. Three studies between 2004 and 2016 used an open-ended questionnaire to test the hypothesis that the association between psychological traits and self-reported well-being is stronger than circumstances and behaviors.
Current measures of well-being, such as the positive and negative affect and satisfaction with life alongside psychological well-being, are robustly correlated with self-reported personality traits and psychological richness. However, this correlation is relatively less robust when considering one’s behavior and circumstances.
Self-reported psychological traits are rated similarly as measures of well-being. Owing to their strong associations with objective measures, such as circumstances, including self-reports to measure well-being earlier seemed justified.
About the study
In the present study, researchers analyzed three samples and three projects from the Midlife in the United States (MIDUS) study. First, they used verbatim responses to the question – “What do you do to make life go well?”. The aim was to evaluate the importance of psychological traits and socioeconomic circumstances for predicting subjective well-being, self-reported by all the participants.
The study used automated zero-shot classification to score statements about well-being without training on existing survey measures. They used hand-labeling to evaluate this scoring later. During follow-up, it helped the researchers assess associations between closed-ended measures for health-related behaviors, socioeconomic circumstances, inflammation and glycemic control biomarkers, and mortality risk.
The team expanded the analyses to cover a 42-item scale that assessed the “psychological well-being” and showed how the survey participants generally felt during the study.
Because of their measurement advantage in close-ended questionnaires, psychological traits have emerged as a strong indicator of well-being. However, in a fairer comparison, as made in this study, the researchers noted that circumstances and behavior would matter just as much. Consequently, the authors asserted that well-being and circumstances are as strongly bound as well-being and personality traits.
These observations reflected similarities in survey response styles rather than the strength of these associations outside the survey context. Thus, organizations should consider targeting an individual’s circumstances to improve their well-being.
Furthermore, the researchers pointed out that all current measures of self-reported subjective well-being earlier depended on close-ended questions. After controlling for behaviors and health indicators, these results did not substantially change associations with mortality risk. Thus, this new open-ended approach provided an overview of well-being from the viewpoint of survey respondents and in a subjective manner.
In future surveys, such targeted, open-ended questions could also uncover unique information about survey respondents and their lives, never before assessed via surveys that ask respondents to self-rate themselves.
The authors also argued that behaviors and circumstances are integral to personality traits and well-being, and neither is more or less consequential. Thus, the study results did not favor or disregard the current trend toward customized interventions in various settings, including public health or clinical settings.
The researchers used hand labels to evaluate the open-ended measures uniquely. They intended to create a set of instructions using which humans could reproduce the machine labels or vice versa. In cases where hand and machine labels were not equivalent, they assessed whether the two types of labels replicated comparable associations between personality traits and circumstances. If not, they might help create more refined well-being measures for future work.
According to the authors, neither hand nor machine labels were superior. While machine labels were useful because of their highly replicable nature, hand labels distinguished general ‘feel’ with the sentiment with which each respondent wrote about doing well. Notably, 1,044 responses in the main analysis set were labeled, and researchers did not discuss the labeling with each other.
Closed-ended survey questions appear less reliable than previously established refined measures of well-being. Yet, there should not be an undue emphasis on an association between well-being and personality traits compared to not self-rated objective measures. Over time, however, these text-based well-being measures across varying contexts would be needed to attain similar reliability for general use.
The ongoing research in language modeling would enable uninterrupted and more refined measurement of subjective well-being in the text without depending on trained algorithms reproducing closed-ended survey measures. However, researchers would require high-quality text data for more technological advances in well-being research.