Loading...

Media is loading
 

Paper Type

Complete

Abstract

Social Desirability Response Bias (SDRB) can adversely affect the integrity of political polling by prompting respondents to modify their honest answers in order to conform with social expectations. This critical issue undermines the accuracy of polling data thereby necessitating innovative detection and prediction techniques. This study, grounded in the self-schema model, applies a novel digital behavioral biometric method by analyzing mouse cursor movements of 99 participants to detect and predict SDRB. Our results indicate a significant relationship between SDRB and various digital biometric behaviors, notably extended answering times, broader mouse movements, decreased cursor speeds, and a higher frequency of answer changes. Additionally, the study employs machine learning models that display impressive efficacy in predicting SDRB, achieving an F1-score of nearly 74%. The observed digital biometric patterns associated with SDRB highlight the potential of these metrics as indicators of respondent authenticity in political polling.

Paper Number

1052

Author Connect URL

https://authorconnect.aisnet.org/conferences/AMCIS2024/papers/1052

Comments

SIGHCI

Author Connect Link

Share

COinS
Top 25 Paper Badge
 
Aug 16th, 12:00 AM

Detecting Social Desirability Bias in Polls: A Digital Behavioral Biometric Approach

Social Desirability Response Bias (SDRB) can adversely affect the integrity of political polling by prompting respondents to modify their honest answers in order to conform with social expectations. This critical issue undermines the accuracy of polling data thereby necessitating innovative detection and prediction techniques. This study, grounded in the self-schema model, applies a novel digital behavioral biometric method by analyzing mouse cursor movements of 99 participants to detect and predict SDRB. Our results indicate a significant relationship between SDRB and various digital biometric behaviors, notably extended answering times, broader mouse movements, decreased cursor speeds, and a higher frequency of answer changes. Additionally, the study employs machine learning models that display impressive efficacy in predicting SDRB, achieving an F1-score of nearly 74%. The observed digital biometric patterns associated with SDRB highlight the potential of these metrics as indicators of respondent authenticity in political polling.

When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.