Keywords
Personality, Big Five, Machine Learning, Natural Language Processing, Prediction, LinkedIn
Abstract
LinkedIn profiles are increasingly serving as supplements to or substitutes for traditional resumes. Beyond the explicit information in LinkedIn profiles, research indicates that recruiters and hiring managers can infer additional applicant characteristics. The Big Five personality dimensions are particularly valuable for organizations to glean from these profiles due to their relationship with job performance. Although previous research has attempted to predict personality from LinkedIn, it has severe limitations, including limited practical utility due to relying on manual coding of profiles, inconsistent and largely non-significant findings, and a tendency to overlook the text data within profiles. This study addresses the first issue by developing an automated computer coding process, which significantly reduces profile coding time. The other limitations are tackled by drawing on the Realistic Accuracy Model and literature suggesting natural language contains personality cues to create a more comprehensive prediction model by incorporating the text data of LinkedIn profiles. Machine learning was used to analyze the profiles of 960 employees recruited through CloudResearch Connect and MTurk. Cross-validated and tested on out-of-sample data, the results indicate that all the Big Five personality dimensions can be validly and reliably predicted from LinkedIn profiles when text data is incorporated and analyzed through open-vocabulary approaches, but generally not when text is not included. Additionally, the built models result in fewer subgroup differences than a traditional self-report personality assessment. This research provides a more efficient and accurate approach to predicting personality from LinkedIn profiles. The implications and limitations of the developed approach are discussed.
Completion Date
2024
Semester
Summer
Committee Chair
Su, Shiyang
Degree
Doctor of Philosophy (Ph.D.)
College
College of Sciences
Department
Psychology
Degree Program
Psychology; Industrial and Organizational Track
Format
application/pdf
Identifier
DP0028627
URL
https://purls.library.ucf.edu/go/DP0028627
Language
English
Rights
In copyright
Release Date
August 2029
Length of Campus-only Access
5 years
Access Status
Doctoral Dissertation (Campus-only Access)
Campus Location
Orlando (Main) Campus
STARS Citation
Tavoosi, Saba, "Predicting Personality from LinkedIn Profiles Using Natural Language Processing" (2024). Graduate Thesis and Dissertation 2023-2024. 424.
https://stars.library.ucf.edu/etd2023/424
Accessibility Status
Meets minimum standards for ETDs/HUTs
Restricted to the UCF community until August 2029; it will then be open access.