How to personalize AI systems to build employee engagement in GenAI use

The study involved 300 participants from various professional backgrounds, ages, and levels of familiarity with AI tools. Participants were randomly assigned and completed a validated Big Five personality inventory to assess their traits: Openness, Conscientiousness, Extraversion, Agreeableness, and Neuroticism. This personality assessment was used to group participants for analysis of their reactions to both the playbook guidance and their interaction with ChatGPT.

 

Materials

AI Interaction Playbook: A structured guide designed to help participants use ChatGPT as a mentor for innovation. The playbook included:

Introduction: Overview of how AI can assist in generating, refining, and testing ideas.

Step-by-Step Instructions: Clear guidance on how to use ChatGPT effectively for idea development. The instructions included recommendations for being polite, asking for confirmation, and seeking reassurance during AI interactions.

Personalization Tips: Participants were advised to customize their interaction style with ChatGPT to match their personal preferences. Examples were given on how to adjust the tone and approach of the chatbot to best suit their needs.

Tailored Examples: Case studies on using ChatGPT across different domains (e.g., marketing, product development) to demonstrate its application in various industries.

ChatGPT Platform: Participants used ChatGPT to follow the playbook and engage in the innovation process, generating and refining ideas while adhering to the guidance on polite and reassuring interactions.

 

Procedure

Pre-Test Personality Assessment: Participants completed a Big Five personality test to establish their scores in Openness, Conscientiousness, Extraversion, Agreeableness, and Neuroticism. These scores were later used to analyze how personality traits influenced reactions to both the playbook and the AI’s responses.

 

Playbook Usage and AI Interaction: Over the course of one week, participants used the playbook to interact with ChatGPT. They were instructed to:

 

Develop innovation ideas with ChatGPT’s help, refining and testing these ideas through multiple prompts.

Follow the playbook’s recommendations to engage with ChatGPT politely, ask for confirmation, and seek reassurance when needed.

Customize their interactions according to their preferences while noting their comfort levels with the chatbot’s responses.

Post-Interaction Survey: After completing the task, participants were surveyed on two main aspects:

 

Playbook Guidance: How useful and engaging they found the step-by-step instructions and recommendations provided by the playbook.

Reactions to AI Friendliness and Reassurance: Their feelings about ChatGPT’s friendly and reassuring behavior, and how it affected their overall confidence, engagement, and satisfaction.

Data Analysis: The data from the post-interaction surveys were analyzed by grouping participants based on their Big Five personality traits. Both quantitative (Likert-scale ratings) and qualitative (open-ended feedback) methods were used to assess how personality traits affected the participants’ experiences with the playbook and ChatGPT.

 

Results

Participant Reactions to Playbook Guidance

Across all 300 participants, the playbook received positive feedback for its clarity and structure. Engagement with the playbook was high (M = 4.2, SD = 0.6), with 80% of participants reporting that the step-by-step guidance helped them use ChatGPT more effectively. The personalization tips were particularly appreciated by participants high in Openness and Conscientiousness, who enjoyed the flexibility and ability to adjust the interaction style based on their preferences.

High Openness (n = 73): These participants found the personalization tips especially useful, as they enjoyed experimenting with different ways to interact with ChatGPT. They reported that the playbook allowed for creative exploration and felt it provided enough room to tailor the experience to their liking (M = 4.6, SD = 0.4).

High Conscientiousness (n = 64): Participants with high conscientiousness praised the playbook’s structure and step-by-step guidance. They appreciated the clear instructions and task-focused approach, which made them feel confident in their ability to use ChatGPT (M = 4.5, SD = 0.5).

Low Openness (n = 60): Individuals low in openness found the playbook somewhat challenging, expressing that they preferred more rigid, straightforward instructions rather than the flexible guidelines provided. They were less likely to engage with the customization aspects of the playbook (M = 3.8, SD = 0.7).

Reactions to GPT Being More or Less Friendly and Reassuring

The study found that participants reacted differently to ChatGPT’s friendliness and reassurance depending on their personality traits, especially in response to the playbook’s guidance encouraging politeness, confirmation-seeking, and reassurance.

 

Openness to Experience

High Openness (n = 73): Participants high in openness appreciated ChatGPT’s friendly tone but felt that too much reassurance or confirmation limited their ability to freely explore ideas. While they enjoyed creative interactions, they preferred a balance between reassurance and the freedom to experiment with the AI’s responses (M = 4.5, SD = 0.4).

Low Openness (n = 60): These participants valued ChatGPT’s friendliness and found reassurance useful, particularly because it made them more comfortable with using AI. They were more likely to ask for confirmation and appreciated the AI’s polite and structured responses (M = 4.0, SD = 0.7).

 

Conscientiousness

High Conscientiousness (n = 64): These individuals appreciated the clear and polite tone, especially when it helped them avoid errors. Reassurance was seen as helpful, confirming that they were on the right track. However, they disliked overly friendly responses that distracted from task completion (M = 4.6, SD = 0.4).

Low Conscientiousness (n = 70): Participants with low conscientiousness enjoyed the relaxed, friendly tone and found the reassurance comforting but less necessary. They were more focused on the creative aspects of the interaction rather than strict adherence to task guidelines (M = 3.9, SD = 0.8).

 

Extraversion

High Extraversion (n = 65): Extroverted participants thrived on ChatGPT’s friendliness and conversational tone. They frequently sought reassurance and found the AI’s encouragement motivating. They described the interaction as lively and engaging, enjoying the social-like aspect of their interaction with the AI (M = 4.4, SD = 0.5).

Low Extraversion (n = 62): Introverted participants preferred concise and task-oriented exchanges. While they acknowledged that the friendliness was pleasant, they felt that excessive reassurance was unnecessary. They were more focused on efficiency than building a relationship with the AI (M = 4.0, SD = 0.7).

 

Agreeableness

High Agreeableness (n = 68): Highly agreeable individuals responded positively to ChatGPT’s friendliness and reassurance. They appreciated the AI’s polite tone and found it helpful in validating their ideas. These participants reported the highest satisfaction with the interaction, particularly because the AI made them feel supported (M = 4.5, SD = 0.4).

Low Agreeableness (n = 58): Participants lower in agreeableness expressed a preference for more direct, no-nonsense communication. They found excessive politeness unnecessary and felt that too much reassurance slowed them down. They preferred blunt, straightforward feedback over emotional support (M = 3.7, SD = 0.8).

 

Neuroticism

High Neuroticism (n = 65): Those high in neuroticism greatly benefitted from the AI’s reassurance. They frequently sought confirmation and appreciated the calming, friendly tone, which helped reduce their anxiety and boosted their confidence in decision-making (M = 4.6, SD = 0.5).

Low Neuroticism (n = 73): Participants lower in neuroticism were less reliant on reassurance and found ChatGPT’s friendliness nice but unnecessary. They preferred task-oriented, concise responses and did not require the same level of emotional support as their high-neuroticism counterparts (M = 4.0, SD = 0.6).

 

Dislikes and Challenges

Despite generally high satisfaction with the playbook and ChatGPT interactions, a few participants reported challenges:

Excessive Reassurance: High-openness and low-agreeableness participants felt that too much reassurance was unnecessary and, at times, patronizing. They preferred the AI to be more direct.

Lack of Directness: Low-agreeableness and high-conscientiousness participants reported that the overly friendly tone sometimes delayed task progress, and they would have preferred more blunt feedback to move quickly through the steps.

 

Summary of Findings

The study demonstrated that personality traits significantly influenced how participants reacted to both the playbook’s guidance and ChatGPT’s friendliness and reassurance. Participants high in Openness, Conscientiousness, Extraversion, and Agreeableness appreciated the playbook’s guidance and tailored their interactions to match their preferences. However, those low in Agreeableness and high in Neuroticism had specific preferences for more direct communication or greater reassurance, respectively. The results highlight the importance of tailoring AI interactions to different personality types to optimize user experience and engagement.

 

Top 3 Most Important Points for Summary:

Personalization Based on Personality Traits: Different personality types respond uniquely to AI interactions. Extroverts prefer more engaging, conversational AI, while introverts favor concise, task-oriented responses. Highly agreeable users appreciate supportive feedback, while low-agreeableness individuals prefer direct, no-nonsense communication.

 

Reassurance and Confirmation: Users with high levels of neuroticism benefit from reassurance and validation, as it reduces anxiety and builds confidence. In contrast, low-neuroticism individuals prefer minimal reassurance, focusing on efficiency and task completion.

 

Flexibility vs. Structure: High-openness users enjoy creative, flexible interactions that allow exploration, while low-openness and high-conscientiousness users prefer structured, step-by-step guidance with clear instructions and confirmations to ensure accuracy.

 

Top 3 Recommendations for AI System Designers:

Incorporate Customization Options: Provide users with the ability to adjust the AI’s interaction style. Allow for toggles between creative or structured responses, and let users control the level of friendliness or formality in communication.

Adaptive Feedback and Reassurance: Design the AI to adapt based on user behavior over time. Offer reassurance and emotional support when needed, but adjust dynamically for users who show a preference for more direct and task-focused interactions.

Contextualized Personalization: Use personality assessments or behavioral cues to tailor AI responses in real-time. Extroverted users might receive more conversational prompts, while task-driven users get focused, actionable guidance.