Will aim to maximize customer human capital while ensuring ethical considerations in AI use including explainability and transparency

Hitachi, in collaboration with The University of Tokyo, has developed AI technology for inferring personality traits making use of paralanguage information in an interview such as facial expressions, gestures, and tone of speech. This technology, while giving proper attention to ethical factors in AI use such as explainability and transparency, is able to infer five personality traits known as the “Big Five”*1 based on psychological theory, for assistance in matching personnel to the appropriate organization or work responsibilities. Further development will be carried out, in collaboration with customers in various areas, aimed at expanding the scope of applicability and improving inference accuracy, thereby helping customers maximize their human capital.

The liquidity of today’s labor markets makes it all the more important to assign the right people to the right positions in a company. Due to such problems as a lack of skillful interviewers and inconsistency in personnel assessment, it has become increasingly difficult to match personnel properly. In response, progress is being made in developing technology that uses AI to assist in assessment of personality traits of those being interviewed based on their answers during an interview (language information). The hope is that if such assistive technology makes possible fair and objective assessment, better matching of personnel to organization and work responsibilities will result. There are two main issues with conventional technology, however. One is the need to conduct learning of paired questions and responses in order to analyze language information (difficulty of altering questions). The other is the difficulty of performing assessment associated with knowledge based in the theory of psychological personality traits.

The new AI technology Hitachi has developed is able to infer personality traits from paralanguage information, based on a model consisting of five psychologically established personality trait factors (the Big Five personality traits). It was developed by combining the psychological knowledge of Professor Kaori Karasawa of The University of Tokyo with Hitachi’s communication behavior analysis technology. To ensure explainability and transparency in the AI inference process, a model was built that infers personality traits by rule-based extraction and synthesis of behavioral
indicators, without using machine learning (neural network), for assessing the Big Five personality traits (Figure 1).

In this study, each of the Big Five personality traits were inferred using this technology based on video and audio data from interviews with 98 subjects who consented to be assessed. The results showed a correlation level of 0.3 or above with each of the personality traits obtained using a (psychologically standardized) questionnaire, confirming the ability to make statistically significant inferences in psychological measures. These results demonstrated the possibility of inferring personality traits based on a psychologically established model using paralanguage information in an interview, without using language information.

Hitachi, in collaboration with customers and partners in various areas, will continue with development aimed at expanding the scope of applicability and improving inference accuracy. Hitachi will further endeavor, in accordance with the Hitachi Principles guiding the ethical use of AI,*2 to ensure protection of privacy, fairness, explainability, and transparency, studying how to use the technology with the understanding and consent of both the interviewer and interviewee, so as to help customers maximize their human capital and match personnel to organizations and work responsibilities.

The results of this development project were announced in part at the 23rd Forum on Information Technology (FIT2024) held in Hiroshima from September 4 to 6, 2024.

画像: Figure 1: Technology for inferring personality traits making use of paralanguage information in an interview such as facial expressions, gestures, and tone of speech

Figure 1: Technology for inferring personality traits making use of paralanguage information in an interview such as facial expressions, gestures, and tone of speech 

*1 Five personality traits derived based on the theory in psychology that human character can be explained by five broad dimensions of personality. They consist of the five perspectives (factors) Emotional stability,
Extraversion, Openness, Agreeableness, and Conscientiousness, as described below.
・Emotional stability: Tendency to relax, to control desires and feelings, and to manage stress easily
・Extraversion: Readiness to approach others, enjoyment of being around people, desire to excel
・Openness: Possession of curiosity, amenability to new theories and to society, acceptance of complexity
・Agreeableness: Acceptance of society and communities (prosociality), lack of hostility
・Conscientiousness: Tendency to plan and think carefully before acting

*2 News release of February 22, 2021, “Hitachi Establishes Principles guiding the ethical use of AI in its Social Innovation Business.”

For more information, use the enquiry form below to contact the Research & Development Group, Hitachi, Ltd. Please make sure to include the title of the article.

https://www8.hitachi.co.jp/inquiry/hitachi-ltd/hqrd/news/en/form.jsp

This article is a sponsored article by
''.