A blog-post authored by Luiza Quinn (BSc Psychology Student) and Professor Linda Kaye
There has been a surge in psychological research using social media in relation to personality. It is a fascinating topic and researchers can gather participants and data with ease through social networks like Facebook and Twitter.
Language is one aspect that has been studied, and Park et al. (2015) developed a model for assessing personality using open-vocabulary analysis called language-based assessments (LBAs).
Open-vocabulary methods characterise single words, non-word symbols (such as punctuation), multiword phrases and clusters of semantically related words (this means words that share a similar meaning or are used together in a particular context, such as clustering job, career, office, task, and professional under the title ‘work’).
The researchers aimed to provide evidence that using LBAs will benefit future investigations to assess a person’s personality from their social media language.
How was this done?
The researchers built a predictive model which linked users' personality traits to their language vocabulary. They used a training sample of over 66,000 Facebook users so that the model could practice. They gathered participants and their data legally and ethically using ‘myPersonality’, an application on the Facebook social network in which users can take a series of psychological measures.
They were given access to every status message (which are social network displays of user’s current activities, moods and thoughts available for anyone to see) written by volunteers between January 2009 and November 2011. This totalled over 15 million messages.
Participants also completed measures of personality traits as defined by the NEO-PI-R five factor model - openness to experience, conscientiousness, extraversion, agreeableness, and neuroticism
The researchers also documented some external criteria to test the accuracy of the model. They did this by using informant (friend) reports of the participant’s personality as well as results from the Satisfaction With Life Scale, the Self-Monitoring Scale, the Orpheus Personality Questionnaire, the Pennebaker Inventory of Limbic Languidness (PILL) and the Barratt Impulsiveness Scale
The model changed the user’s status messages into counts of simple language features. These features were:
individual words and phrases
topics.
Then researchers counted how many times the features appeared in each users language sample. Using this data and the external criteria for personality, correlations could be formed
4,824 users were later used as a validation sample after the model training had concluded.
So, what did they find?
Research findings are based around how LBAs may complement and extend traditional measures in social media samples by providing an alternative to self-report questionnaires.
From the 4,824 users, correlations were made between the LBAs and self-reports of the Big Five personality traits (Openness, Conscientiousness, Extraversion, Agreeableness, and Neuroticism) they were found to be positive and substantial.
Overall, the LBAs provided a broad coverage of the personality domains.
On average, LBAs were similar in agreement (or accuracy) with the external criteria
Implications for this research
Compared with self-report questionnaires, LBAs are extremely fast. The entire validation sample, roughly 5,000 participants, was assessed in minutes.
In this study, the researchers used language to assess Big Five personality traits, but LBAs are not limited to personality. This same method can be adapted to create language models of other psychological characteristics, including psychological well-being, attitudes, traits in other personality frameworks (e.g., HEXACO; Ashton & Lee, 2007), and more temporary states such as mood
Questionnaires can be expensive to administer and time and resource intensive. LBAs offer a practical, cost-effective alternative, allowing assessment of psychological characteristics when questionnaires are impractical.
References
Park, G., Schwartz, H. A., Eichstaedt, J. C., Kern, M. L., Kosinski, M., Stillwell, D. J., ... & Seligman, M. E. (2015). Automatic personality assessment through social media language. Journal of Personality and Social Psychology, 108(6), 934. https://doi.org/10.1037/pspp0000020
Comments