In-Between Visuals and Visible: The Impacts of Text-to-Image Generative AI Tools on Digital Image-making Practices in the Global South

Title of the Paper

"In-Between Visuals and Visible: The Impacts of Text-to-Image Generative AI Tools on Digital Image-making Practices in the Global South"

Paper Information

• Subject Area: Human-Computer Interaction (HCI), Generative AI, Digital Image-making, and Cultural Impacts
• Keywords: Generative AI, Text-to-Image, Human-Computer Interaction, Data Colonialism, Global South, Digital Creativity, Inclusive Design, Image Generation Models, Data Bias, Postcolonial Computing

Research Background and Issues

• Identified Problems or Challenges:

  • Existing generative AI (GAI) tools have profound cultural and economic impacts on the Global South (e.g., Bangladesh), but their specific effects remain underexplored systematically.
  • Text-to-image generative models often produce inaccuracies in cultural contexts, leading to exclusion and distortion of local cultures.
  • Local image creators face challenges related to language barriers, lack of data, and Westernized visual narrative styles, limiting their creative capacity and marginalizing their artistic practices.

• Significance of the Research:

  • Generative AI is rapidly transforming the creative work domain, influencing cultural identity, as well as the economic and creative practices of local creators in the Global South.
  • Investigating whether generative AI exhibits data colonialism and how it affects cultural representation is crucial for promoting social equity, decolonization, and global development goals within HCI.

• Motivation and Related Work:

  • Existing studies primarily focus on ethical risks and technical limitations of GAI, predominantly in Western contexts, with limited exploration of its specific impacts on cultural and artistic practices in the Global South.
  • This paper aims to address this research gap by uncovering issues of social and cultural exclusion in Bangladesh's digital image industry caused by GA tools, offering solutions through the lens of postcolonial computing.

Solutions

• Proposed Methods and Solutions:

  • Conducted a five-month qualitative field study involving in-depth interviews, focus group discussions (FGD), and social media analysis with image creators in Bangladesh to systematically explore the impacts of GAI tools on image design.
  • Analyzed the influence of generative AI tools on various stages of image production (creation, depiction, presentation) from cultural, linguistic, and design perspectives.

• Innovative Contributions:

  • Introduced the concept of "image-mediated colonialism" to describe how generative AI extends postcolonial impacts through exclusion and distortion in cultural representation.
  • Explored solutions from social equity and design perspectives, such as developing AI systems with heightened language sensitivity and enhancing the visibility and networking of local content.

• Implementation Steps and Key Techniques:

  • Analyzed the image generation functionalities of different generative AI models (e.g., Midjourney, DALL-E) and their applicability in the Bangladeshi context.
  • Employed semantic analysis and social research techniques to interpret linguistic barriers and cultural inaccuracies in AI creation processes.
  • Proposed strengthening AI’s cultural sensitivity through localized databases and text-image alignment models.

Research Findings

• Specific Findings:

  • Developed a three-stage framework (creation stage, depiction stage, presentation stage) to deeply analyze how GAI technologies impact the image-making process in Bangladesh.
  • Highlighted major issues in text prompt generation, exclusion and distortion of cultural symbols, and lack of sensitivity to social ethics in GAI systems.

• Comparison with Existing Solutions and Advantages:

  • Unlike studies dominated by Western perspectives, this paper provides field data and case descriptions from the Global South, emphasizing the importance of cultural context.
  • Critically analyzed the impacts of data colonialism, language barriers, and accelerated market economies on image-making practices, reexamining the dynamic relationship between artistic creation and technological development.

• Experimental or Evaluation Results:

  • Investigations revealed that generative AI has become a key tool in Bangladesh but has marginalized many small-scale creators due to language and cultural limitations.
  • Comparative data showed that designs by local architect Bashirul Haq were significantly misunderstood or overlooked by GAI systems, unlike those of Western architects such as Zaha Hadid.

• Limitations and Future Directions:

  • Limitations: The study primarily focused on urban areas, with limited exploration of rural image-making practices; traditional handcrafted image-making was not extensively included.
  • Future Directions: Investigate how policy and design interventions can enhance technological access and language support for creative industries in the Global South; further explore GAI impacts in rural areas and broader Southern countries.

This study provides valuable case studies and theoretical perspectives for promoting diversity, decolonization, and equitable design, offering critical reflections on generative AI from the standpoint of Global South practices.