Digital Inclusion for Social Good

One key issue with datasets becoming integral to daily life is algorithmic exclusion and inherent bias. Algorithms exclude and perpetuate marginalised populations, cultures, and ways of being in ways that reflect the social relations prevalent online. This creates power asymmetries over control and use of data, with power concentrated in a few companies with limited public accountability. Low digital literacy and diminishing community agency increase this problem, as do gaps in managing online risk and data safety.

The field observations, as illustrated below, are based on a series of qualitative studies with various demographics conducted in 2025-26 in smaller towns in India, with particular focus on women using digital content through personalised apps; gig workers associated with platforms; and people from marginalised backgrounds.

Gender remains a persistent and consequential axis of digital inequality. Women Participants described lived experiences of stigma, exclusion, and discrimination, especially around gender. This highlights why feminist approaches

must address how patriarchy, asymmetries, and other structures act together to produce inequality.

The gig economy centres task-based, piecework assignments with short-term contracts and flexible timings, mediated by digital platforms. Evaluating gig livelihoods requires attention to platform indifference and the absence of worker protections.

Digital divide and intersectional exclusion: Socioeconomic barriers prevent many marginalised individuals from accessing reliable technology or internet connections, particularly in small-tier towns. All stakeholders must rethink the adoption of emerging technologies not only by capability but also through the lenses of power, accountability, and sustainability.

Exclusion in Tech Design & Decision-Making

Beyond work-related and educational use, digital technologies are quietly reshaping social

practices. The digital divide and simultaneous surge of Artificial Intelligence and Large language models in everyday life have kept exclusion deeper, which should not be a binary of inclusion versus exclusion but larger economic structures.

Factors Keeping Marginalised Outside the Tech Design & Decision-Making

• Lack of diverse voices at the design table— Communities absent; systems don’t serve unique needs
• Insufficient data from marginalised populations — creates blind spots in model performance
• Historical exclusion from tech development— corporations/governments excluded groups from deployment/oversight
• Concentration of AI infrastructure— infrastructure concentrated in Global North; South excluded from rule-setting
• Data extracted from Global South without consent; economic value accrues to Global North companies
• Unclear transparency— ‘black box’ models limit accountability for bias correction

Cases and regional examples

• Asia: Researchers and civil-society groups in South Asia and Southeast Asia have documented LLM outputs that reinforce occupational and cultural stereotypes about women and certain ethnic groups.
• Africa: Studies from African researchers and NGOs show that many LLMs trained primarily on Western text produce inaccurate or biased characterisations of African women and underrepresent local languages and gender identities, leading to misgendering and exclusion in local languages.

Latin America: Researchers in Latin America have reported that AI systems sometimes replicate patriarchal language patterns and stereotypes, and regional advocacy has pushed for gender-inclusive corpora and participatory evaluation with local feminist organisations.