The Neurophysiological Paradox of AI-Induced Frustration: A Multimodal Study of Heart Rate Variability, Affective Responses, and Creative Output

Figures

• 

  Figure 1. Experimental procedure. Solid arrows represent the temporal flow of tasks; the gray dashed arrow indicates the transition between Day 1 and Day 2. HRV = Heart Rate Variability. PANAS = Positive and Negative Affect Schedule. CSES = Creative Self-Efficacy Scale. NASA-TLX = NASA Task Load Index. Time is measured in minutes (min).

• 

  Figure 2. Boxplot of changes in positive (PA) and negative (NA) affect for the control group (A) and the AI-assisted group (B). Values represent post–pre difference scores. The gray dashed line indicates no change from baseline. The horizontal line inside each box indicates the median; boxes represent the interquartile range (IQR), whiskers extend to 1.5 × IQR. Diamonds represent outlier values beyond this range.

• 

  Figure 3. Boxplot comparison of subjective task load ratings across six NASA-TLX dimensions between the control group (A) and the AI-assisted group (B). Each box shows the interquartile range (IQR), the line inside the box represents the median, and whiskers extend to 1.5 times the IQR. Diamonds represent outlier values beyond this range. The color coding of groups has been revised to ensure consistency with the legend.

• 

  Figure 4. Group-wise HRV changes across task phases. (a) Line plot of LF/HF ratio across task phases (T1–T3), with separate trends for the control group and the AI-assisted group. (b) Line plot of RMSSD values across the three phases, showing a sharper decline in the AI-assisted group. Error bars denote ±1 SD.

• 

  Figure 5. Scatterplot and correlation analysis of HRV physiological indicators (RMSSD vs. LF/HF) with subjective variables (negative mood, effort, and frustration). Regression lines with 95% CI are shown.