24042236's blog

Intervention 2.2 ： Data Analysis & Reflection

—

by

Zhihui Li
in

https://24042236.myblog.arts.ac.uk/files/2025/08/AI视觉内容与情感链接测试AIVisualsEmotionalConnectionTest报告2.docx

1. Sample Overview

  • Sample size: 39 participants (all consented).
  • Age distribution: Majority aged 26–35 (48.7%), followed by 36–50 (28.2%), with younger participants aged 18–25 accounting for 15.4%.
  • Gender: Male-dominated (61.5%), female 25.6%, non-binary/other 2.6%, undisclosed 10.3%.
  • Fashion interest: “Occasional” interest was most common (43.6%), “frequent” and “rare” interest each 25.6%, with only 5.1% reporting “no interest.”

The sample was relatively diverse but skewed toward young to middle-aged men with moderate fashion interest.

2. Key Findings

Brand Representativeness (Q5)

  • Image A was perceived as most representative of Chanel (41%), followed by Image B (28%) and Image C (26%).
  • This indicates clear perceptual differences in “brand congruence” across AI/real combinations.

Emotional Appeal (Q6)

  • Image A (49%) was considered most emotionally appealing, followed by Image B (36%).
  • Image C was rated much weaker (8%).

Four Dimensions (Q7–Q10)

  • Warmth: Overall mean 4.98, with A (5.41) > B (5.03) > C (4.51).
  • Emotional connection: Overall mean 4.85, A (5.10) slightly above B (4.87), C lowest (4.59).
  • Realism: Overall mean 5.07, with A (5.41) and B (5.05) higher, C lower (4.74).
  • Brand congruence: Overall mean 4.92, A (5.28) > B (4.97) > C (4.51).

Image A consistently led across all four dimensions, particularly realism and warmth.

Attitude Change After Disclosure (Q12)

  • 57% reported they would change their choice (mostly shifting toward A or B), while 43% would not.
  • Indicates AI usage transparency prompts consumer reconsideration.

Ranking of Influencing Factors (Q14)

  • Most important: Presence of real models (51.6% ranked first).
  • Second: AI usage transparency (44.4% ranked first).
  • Third: Over-perfection / eeriness (52.4% ranked first, but lowest overall scores).

Consumers cared most about real models and transparency.

Acceptable AI Applications (Q15)

  • Highest acceptance: background (72%), colour/composition enhancement (67%), props/apparel (62%).
  • Only 38% accepted AI replacing human models.

Consumers tend to accept AI in supporting roles but not in replacing central human figures.

3. Analysis & Interpretation

Lack of Emotional Resonance

  • Although realism was rated high (>5), warmth and emotional resonance were consistently lower (<5).
  • Suggests AI images can “fool the eye” but struggle to “move the heart.”

Transparency as a Double-Edged Sword

  • Disclosure of AI usage changed some choices, showing transparency builds rational trust.
  • Yet, over-emphasis on AI may lower perceived warmth, creating a “cold” impression.

Irreplaceable Role of Human Models

  • Real human presence was ranked most important, aligning with fashion’s reliance on people as emotional mediators.
  • Explains why replacing models with AI was least acceptable.

Clear Boundaries of Application

  • AI is widely accepted for backgrounds, props, and retouching.
  • But once applied to core figures, trust and emotional connection significantly decline.

4. Value & Limitations

Value

  • Multidimensional perception: Shows that realism ≠ emotional resonance.
  • Boundary recognition: Identifies AI’s “safe zones” (background/props) versus “sensitive zones” (model replacement).
  • Transparency tension: Highlights the delicate balance between disclosure and emotional connection.
  • Attitude shift: Demonstrates a dynamic path from intuitive choice → AI disclosure → changed attitude.
  • Practical relevance: Results can be translated into brand AI usage boundary guidelines, providing actionable frameworks for industry.

Limitations

  • Small sample size (N=39), skewed toward young/middle-aged men, limits representativeness.
  • Relies on intuitive survey responses; lacks longitudinal or qualitative depth to explore “why.”
  • Context restricted to Chanel, limiting generalisability to other brands.

Conclusion

This experiment shows that while AI visuals perform well in terms of realism, they remain weaker in emotional resonance and brand congruence. Consumers broadly accept AI in supporting roles, but resist its use in replacing human figures.

References (Harvard Style)

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *