Decisions Lab CEO Louis Cheung presented at the Design Thinking in Action (DTIA) Get-Together forum held recently in Singapore, highlighting a critical yet often overlooked aspect of AI: the inherent cultural bias within large language models (LLMs).
Cultural Bias: An Inevitable Reality
Louis demonstrated through live interactions how prominent LLMs—ChatGPT, Claude, and DeepSeek—each responded differently to identical prompts, clearly showcasing unique cultural "accents." Research consistently indicates that mainstream models such as ChatGPT reflect predominantly Western-centric norms, whereas emerging models like DeepSeek have a noticeable pro-Chinese inclination, particularly on sensitive subjects (OUP Academic, Medium).
The session underlined Decisions Lab’s perspective: Bias is not a flaw but a reflection of training data and cultural contexts. The crucial responsibility lies in consciously acknowledging and strategically managing this bias.
Prioritizing Empathy in AI Design
Central to DTIA's discussions was the significance of empathetic design, illustrated through cases from elder-care facilities and marketplace platforms. Each scenario emphasized a core principle: the power of context in effective solution-building.
Louis connected these insights to Decisions Lab's expertise in AI simulation, where thousands of virtual personas undergo scenario testing to identify cultural and contextual blind spots. Highlighting a real-world application, Louis cited the effectiveness of locally tuned retention marketing campaigns—those culturally aligned to specific audiences outperform generic, culturally mismatched content.
Key Insights for Effective AI Development
Participants at DTIA gained clarity from three pivotal insights shared by Decisions Lab:
- Context Over Capability: Even advanced AI models falter when misaligned with local cultural nuances.
- Human-Centric Decision-Making: AI provides insights and drafts, but designers ultimately guide its ethical and practical application.
- Localization is Crucial: Effective AI implementation demands recognition and adaptation to diverse cultural identities, particularly across Asian markets.
Future of Responsible AI
Louis concluded by challenging the conventional approach to AI bias. True responsibility, he argued, lies not in the unrealistic goal of removing biases but in transparently mapping and strategically steering these biases to align with desired outcomes—whether in healthcare, marketplace integrity, or enhancing retention marketing campaigns.
With DTIA scheduled to convene next year in Hong Kong, Decisions Lab remains committed to advancing culturally responsive AI solutions, continuing to bridge global innovation with nuanced local execution.