Research

The academic and professional foundation behind Default vs. Designed.

  • Western, English-dominant training data produces outputs that default to white, Western representations of success, beauty, and professionalism. This is not a glitch. It is the system working as designed.

  • Survey data collected for this project found 50% of respondents would trust a brand less after learning AI made their ads. Case studies from Aerie, Heineken, and Coca-Cola confirm the backlash is measurable and public.

  • Hofstede's six dimensions and Hall's high and low context model provide the analytical backbone for understanding where and how AI defaults break down across different audiences.

  • AR is not decoration here. Research on embodied learning supports the use of immersive media as a tool for shifting perspective in ways static comparison cannot achieve.

Overview

This project sits at the intersection of three fields: AI and bias, cross-cultural communication, and immersive media. The research below documents the frameworks, case studies, and academic sources that informed every design decision in the exhibition.

What the Data Shows

When asked how they would feel about a brand after learning AI made its ads, 50% of respondents said their trust would decrease. Only 3% said it would increase. The gap is not a perception problem. It reflects a values misalignment between how the industry is adopting AI and what audiences actually expect from the brands they trust.

n = respondents to primary research survey conducted February 2026

Primary Research

A quantitative survey was conducted as part of this project to measure audience attitudes toward AI-generated advertising. Distributed digitally, the survey collected structured responses on trust, cultural awareness, and perception of AI in branding contexts.

These findings suggest a measurable gap between industry adoption of AI creative tools and consumer trust in the outputs they produce.

Sources

AI & Systemic Bias

  • Ferrara, E. (2024). Fairness and bias in artificial intelligence: A brief survey of sources, impacts, and mitigation strategies. MDPI.Link

  • Brookings Institution. (2023). Rendering misrepresentation: Diversity failures in AI image generation. Link

  • Zhou, M., & Abhishek, V. (2024). Bias in generative AI. arXiv.Link

  • OpenAI. (2023). DALL-E 3 system card. Link

Consumer Trust & Authenticity

  • Schilke, O., & Reimann, M. (2025). The transparency dilemma: How AI disclosure erodes trust. Link

Brand Strategy & Industry Response

  • Business Insider. (2025, October). The hot new trend in marketing: Hating on AI. Link

  • Business Insider. (2025, October). Aerie's promise not to use AI in ads is its most popular Instagram post in a year. Link

  • Business Insider. (2024, November). This year's Coca-Cola holiday ad exposes one of the biggest problems with AI-generated video. Link

Cross-Cultural Communication

  • Hofstede, G. (n.d.). The 6-D model of national culture. Hofstede Centre.Link

  • Sheposh, R. (2023). High-context and low-context cultures. EBSCO Research Starters.Link

AR Pedagogy & Embodied Learning

  • Bailenson, Jeremy N. (2018). Experience on demand: What virtual reality is, how it works, and what it can do. W.W. Norton & Company.

AI Tools: Use and Disclosure

This project used AI tools across research, design, and production. The following documents where and how each was used.

  • ChatGPT (OpenAI) Used to generate branding copy and imagery as part of the core methodology. Prompts were intentionally submitted without cultural context to surface default outputs. Those outputs were documented as is, not edited or improved.

  • Sora (OpenAI) Used to generate video content for the website landing page.

  • Gemini (Google) Used in early image generation testing before Ideogram was identified as the stronger tool for this project's visual needs.

  • Ideogram Used to generate the redesigned imagery for the AR poster experiences, the intentionally designed side of the Default vs. Designed comparisons.

  • Claude (Anthropic) Used for copy drafting and editorial support throughout the development of this project.

  • ZapWorks Designer Not a generative AI tool, but the immersive design platform used to build and publish the AR poster experiences.