Computational Personality Assessment - An Overview and Perspective

Abstract

Computational methods for the representation and analysis of data have drastically increased the objectivity, reliability, and the practical implications of research conducted throughout most scientific pursuits. Our rapidly-emerging potential to transform digital data into objective measures of human behavior, thoughts, and feelings has perfectly positioned personality science as a critical discipline that will benefit from today’s ongoing digital revolution. Here, we review and discuss some of the most promising approaches to computational personality assessment based on data from experience sampling, natural language, online social media, mobile sensing, and images. We present a concise overview of key findings, discuss the potential and promise ofcomputational personality assessment, and highlight important remaining questions in their development and application. We conclude with an optimistic outlook on how computational assessment could fuel the transition from personality research to personality science.

Applicant Reactions to Digital Selection Methods: A Signaling Perspective on Innovativeness and Procedural Justice

“…Traditional assessment methods with a low degree of digitalization include work sample tests conducted on the premises of the organization, written (paper-andpencil) cognitive ability tests, personality tests, situational judgment tests, or assessment centers (Macan et al, 1994;Ryan & Ployhart, 2000;Steiner & Gilliland, 1996). On the other hand, online work sample simulations (Tippins, 2015), gamified online assessments (Armstrong et al, 2016;Buil et al, 2020), web-based cognitive ability tests (Potosky & Bobko, 2004), computational personality assessments (Stachl et al, 2021), or online-based situational judgment tests (Woods et al, 2020) can be considered assessment methods with a high degree of digitalization. Traditional structured or unstructured face-to-face interviews (Smither et al, 1993) are selection methods used at the interview stage with the lowest degree of digitalization.…”

The computational turn in online mental health research: A systematic review

“…Ethical considerations are one of the most discussed issues in the academic debate around SM data and computational methods (for an overview, see Salganik, 2019). Current open questions include the distinction between public and private content (McKee, 2013), the obfuscating of sensitive and personally identifiable information, data de-identification, algorithmic inferences, dealing with deleted or removed data (Chancellor et al, 2019b), users revealing sensitive personal information about others (Bagrow et al, 2019), implanting meaningful protocols for gaining consent (Chancellor et al, 2019a) and integrating individuals' perspectives, as well as holding users accountable for their past actions (Stachl et al, 2021). Furthermore, the evaluation of online MH is a sensitive area that can be isolating and stigmatizing, putting vulnerable populations and online communities at risk of harm (Chancellor et al, 2019a(Chancellor et al, , 2019bKalluri, 2020).…”

“…Due to the beforementioned challenges of questionnaire-based self-reports, scholars are working to make use of the great variety of digital traces on SM (Stachl et al, 2021), aiming to enable more accurate 'microscopes' of individual human behaviour, as well as 'macroscopes' for collective phenomena (Correia et al, 2020). Most prominent are text features (Sinnenberg et al, 2016), specifically user comments, which are an important form of SM data across disciplines (Schindler and Domahidi, 2021).…”

Personality computing: New frontiers in personality assessment

“…For example, if the goal is to replace traditional personality trait measures, performance should at least approach self‐other ( r s ≈ 0.29–0.41; Connelly & Ones, 2010) or between‐inventories correlations ( r s ≈ 0.31–0.56; Pace & Brannick, 2010). But if the goal is to complement (rather than replace) traditional assessment, then even with lower performances PC models can yield valuable insights into which behavioral variables are indicative of traits (Stachl et al., 2021). Second, the lack of meta‐analytic aggregation prevents conclusive evaluations of PC performances and their heterogeneity.…”