| dc.description.abstract | This study focused on AI induced anxiety, specifically how anxiety levels are affected when interacting with AI that employs prompt engineering techniques. Most of the research being done with AI is conducted to determine what prompt engineering techniques produce more accurate results, reduce hallucinations, and minimize misinformation. For that reason, it is necessary to study the effects of prompt engineering on users.
This research examined how prompt engineering, neuroticism, and contextual setting influence the user's anxiety levels in each AI interaction. I studied different prompt engineering techniques and language that affects interactions with neurotic individuals to determine the methods that would be most impactful in altering anxiety levels of neurotic individuals. To employ the prompt engineering techniques described in my hypotheses, I wrote custom code to assist with the generation of different prompts: each with different personas, contexts, and instructions that were then fed into an LLM to produce responses. Responses were then used in a survey of 996 TCU students to determine if anxiety levels changed based on the AI interaction.
Most of the hypotheses concerning the effects of different prompt engineering techniques on anxiety levels were supported; however, the vast majority of respondents, regardless of neuroticism levels, felt more anxious when faced with responses from bad prompt engineering techniques. This study concluded that neuroticism does not solely affect users' anxiety levels when interacting with AI, it revealed that prompt engineering techniques and contextual settings have a stronger influence over AI-anxiety than neuroticism. The next step would be to conduct research on the levels of anxiety normally felt with a human interaction in the same contexts as were used in this study's AI interaction to determine how influential AI is over anxiety levels. | |
