How Culture Shapes What People Want From AI

Title of the Paper

How Culture Shapes What People Want From AI

Bibliographic Information

• Field of Study: Human-Computer Interaction (HCI), Artificial Intelligence, and Cultural Studies
• Keywords: Culture, Independence/Interdependence, Agency Model, Equality, Diversity, Human-Centered AI, Theory, Survey Research

Research Background and Problem

• Issues and Challenges: The design and development of artificial intelligence are often dominated by Western cultural perspectives (particularly the independence model), neglecting the impact of global cultural diversity on expectations of AI. This limitation may result in failing to meet the actual needs of diverse cultural groups and restrict the potential of AI design.
• Significance: Human society is culturally diverse, yet current AI theories and practices, centered around "Western, Educated, Industrialized, Rich, and Democratic (WEIRD)" cultures, underestimate the influence of cultural diversity on expectations and usage patterns of AI.
• Research Motivation: The authors aim to propose a more inclusive and culturally responsive theoretical framework for human-computer interaction design by exploring how culture influences people's expectations of AI forms and functions. This study also seeks to understand the cultural differences between the U.S. and China regarding preferences for AI characteristics.

Solution

• Proposed Model and Theoretical Framework:

  • The core model guiding this study is the "Independence and Interdependence Cultural Model," which emphasizes how individuals' relationships with their environment shape their expectations of AI characteristics based on cultural contexts.
  • Independence Cultural Model (commonly seen in European Americans): Individuals primarily view themselves as independent from social and physical environments, expecting greater control over AI and placing lower demands on AI autonomy and influence.
  • Interdependence Cultural Model (commonly seen in Chinese individuals): Individuals tend to perceive themselves as closely connected to their environment and others, expecting to establish a closer connection with AI and being more inclined to accept highly agentic AI (e.g., AI with proactive influence).
  • Hybrid Cultural Model (commonly seen in African Americans): Combines characteristics of both independence and interdependence, showing preferences that fall between the two.

• Methodology: Two survey studies were conducted:

  1. Pilot Study: Investigated the expectations of three cultural groups (European Americans, African Americans, and Chinese) regarding "ideal human-environment relationships."
  2. Main Study: Explored how these cultural differences map onto preferences for "ideal AI-human interaction characteristics."

• Innovations:

  • Proposed an AI design framework based on cultural psychology theories, broadening the current AI design perspective, which is primarily dominated by the independence cultural model.
  • Developed new scales to measure "ideal models of human-environment relationships" and "desired characteristics of AI influence."

Research Findings

• Key Findings:

  1. Impact of Culture on Human-Environment Relationship Models:
     • European Americans tend to emphasize "humans influencing the environment," while Chinese individuals lean towards "the environment influencing humans." African Americans fall between these two perspectives.
  2. Impact of Culture on the Need for Control and Connection:
     • European Americans focus more on control over AI, whereas Chinese individuals place greater importance on a sense of connection with AI. African Americans show higher connection needs than European Americans but no significant difference in control needs.
  3. Cultural Preferences for AI Influence Characteristics:
     • Chinese individuals are more supportive of AI possessing higher levels of autonomy, emotionality, and engagement in social interactions. European Americans exhibit relatively lower support for these characteristics, with African Americans positioned between the two groups.

• Strengths:

  • Theoretical Contribution: Provides a cultural perspective on AI preferences, enriching the HCI field and encouraging designers and developers to reflect on implicit cultural assumptions.
  • Practical Implications: Promotes AI design methods that are more globally inclusive through a cultural diversity framework.

• Experimental and Evaluation Results:

  • All hypotheses were validated through data analysis in the main study.
  • Further moderation analysis revealed that cultural background influences preferences for AI influence characteristics through the mediating variables of "control" and "connection."

• Limitations and Future Directions:

  1. Sample Representativeness: The sample size is relatively small, and there are demographic differences (e.g., age, income) among the three groups.
  2. Measurement Tool Improvement: The newly developed scales require further validation of reliability and validity in other contexts and populations.
  3. Specific Application Scenarios: The current study focuses on general conclusions; future research should explore cultural preferences in specific AI system designs.
  4. Behavioral Studies: Beyond surveying preferences, future research should design interaction experiments to study users' actual behaviors when interacting with AI systems.
  5. Expanding Cultural Scope: Investigate more diverse samples (e.g., other countries and religious groups) to uncover richer cultural differences.

Conclusion

This study proposes a theoretical framework for understanding how cultural background influences people's preferences for AI and the underlying mechanisms. Through empirical research, it confirms the significant impact of culture (independence/interdependence models) on AI design preferences. Future research should expand and deepen this direction by incorporating more diverse samples, specific scenario designs, and real interaction methods to provide guidance for achieving more inclusive and culturally responsive human-computer interaction design.