Vidushi Marda

Vidushi Marda is an Indian human rights lawyer and researcher specializing in the societal consequences of AI systems in non-Western contexts, and the corpus's primary voice on AI harms documentation in the majority world — bridging the academic AI fairness-and-accountability research community, international civil society advocacy, and international human rights law (see Person entry). She established and led ARTICLE 19's inaugural international AI programme as Digital Programme Officer on Algorithmic Decision Making, co-authoring the foundational 2018 joint report on privacy and freedom of expression in the age of AI with Privacy International, and went on to produce the first ethnographic analysis of predictive policing in India and the first systematic investigation of China's emotion recognition market — research that was cited by India's Supreme Court, the UK House of Lords, and the UN Special Rapporteur on freedom of opinion and expression, and carried by The Guardian, Reuters, Financial Times, and Politico. She co-founded REAL ML (Research, Engagement, Advocacy & Lived Experience in Machine Learning), a global interdisciplinary practitioner network for algorithmic accountability, and currently serves as Director of Ecosystem Building at The Collaborative for AI in the Public Interest, where she leads work on worker power across the AI supply chain.

The Voice anchors three movement-area registers that the corpus's voices slice had previously left empty.

• The majority-world AI research register. The corpus's existing AI research voices — Joy Buolamwini, Cathy O'Neil, Timnit Gebru — anchor the algorithmic-accountability register primarily through US-based research institutions and the FAccT conference community. Marda occupies a structurally distinct position: the India-based researcher who documents how AI systems actually function in non-Western institutional contexts — the Delhi Police predictive policing system, the Chinese emotion recognition market, Indian facial recognition procurement — where the social conditions, legal frameworks, and power asymmetries differ structurally from the Global North settings where most AI governance research is produced. Her 2020 FAT* paper and 2021 Nature Machine Intelligence article on ethnographic methods are the corpus's clearest academic anchors on this methodological and geographic gap in AI research.

• The "governance with teeth" argument — the ARTICLE 19 AI programme register. The corpus tracks ARTICLE 19 as the London-based international human rights organization that anchors the freedom-of-expression / AI-governance intersection in civil society advocacy. Before this entry, that institutional anchor had no corresponding Voice. Marda is the on-record architect and public voice of ARTICLE 19's inaugural international AI programme — the programme through which the "governance with teeth" framing was developed: the argument that AI ethics principles and FAT initiatives, however technically sophisticated, lack the legal enforceability and mandatory accountability mechanisms that human rights law provides. The voice closes the ARTICLE 19 AI programme anchor entirely.

• The India-based digital rights / AI Voice anchor. The corpus's two existing India-based voices — Anita Gurumurthy (IT for Change, feminist data governance, AI constitutionalism, UN multilateral engagement) and Apar Gupta (Internet Freedom Foundation, civil-liberties litigation, Indian regulatory context) — occupy the scholar-practitioner-in-multilateral-processes register and the civil-liberties-lawyer-founder register respectively. Marda's register is structurally distinct: the empirical researcher who produces peer-reviewed work in AI fairness conferences (FAT*/FAccT, ACM), argues for ethnographic methods as a methodological discipline, documents AI systems in Indian and Chinese contexts, and carries those findings from academic venues into civil society reports, Indian op-ed media, and international policy processes. The three voices are complementary anchors of the Indian AI-governance landscape, not substitutes.

Key research contributions

Marda's primary public output runs through four overlapping bodies of work.

• Predictive policing in India — the first ethnographic account. The "Data in New Delhi's Predictive Policing System" paper (co-authored with Shyam Narayan, FAT* '20, January 2020) is the corpus's primary anchor on how predictive policing functions in a non-Western context. The analysis documents how Delhi Police's crime prediction system operationalizes historical arrest data — data produced by prior biased policing — rather than independent crime signals, and argues that the system's apparent technical legitimacy actively conceals the social structures that produced its training data. Published at the same time as her Nature Machine Intelligence article (Vol. 3, 2021, with Narayan) arguing that ethnographic methods are indispensable to AI research, the Delhi policing work is the corpus's clearest single artefact of that methodological argument: the system's harms were only visible through on-the-ground research, not through technical auditing alone.

• China's emotion recognition market — the first systematic investigation. The "Emotional Entanglement" report (co-authored with Shakir Ahmed, ARTICLE 19, January 2021) is the corpus's primary source on how China's emotion recognition industry intersects with surveillance and human rights. The report documents the market's scale, its deployment in workplaces, schools, and public spaces, and the human rights implications of systems that claim to read emotional states from facial images — technology whose scientific basis is contested but whose commercial and governmental adoption is rapid. The report carried into The Guardian, Reuters, Financial Times, Politico, and Fast Company as international mainstream media coverage of the field — anchoring the public narrative on emotion recognition risk through Marda's and Ahmed's systematic evidence base.

• The "Governance with Teeth" argument. The ARTICLE 19 "Governance with Teeth" report (April 2019) is the corpus's clearest single artefact of the human-rights-vs.-ethics argument in AI governance. The report articulates the structural gap: AI ethics guidelines and the FAT research community's accountability frameworks lack the enforcement mechanisms, mandatory due diligence requirements, and independent accountability structures that human rights law already provides in other technology domains. "Governance with teeth" names what ethics principles cannot achieve on their own — not a supplementary framework but a replacement of voluntary commitments with binding legal obligations. The argument recurs across Marda's FAT* '20 paper "Leap of FATE" (co-authored with Cath, Latonero, and Pakzad) and her EDRi-co-authored piece "Emotion (mis)recognition: Is the EU Missing the Point?" (with Ella Jakubowska, February 2023), which extends the argument to the EU AI Act's treatment of emotion recognition.

• Indian mainstream media and policy engagement. Marda's op-ed register across Indian media ran her research on facial recognition, predictive policing, and biometrics into public discourse before the policy debates that followed. Her op-eds in The Hindu, Hindustan Times, India Today, Times of India, and The Print through 2019–2021 documented the expansion of facial recognition into Indian law enforcement and public space, named the civil liberties risks, and directly fed the policy debate surrounding India's National Automated Facial Recognition System (NATFRS) procurement. The parallel policy engagement channel — presenting at the UN Human Rights Council, briefing diplomats ahead of resolution negotiations, advising on biometrics during the EU AI Act process, and serving as a civil society representative at the Bletchley AI Safety Summit (November 2023) — ran the same research findings into international regulatory and treaty processes.

Signature framings

Three formulations recur across Marda's public output and have done the most to install her register into the AI governance and algorithmic accountability fields.

• "Governance with teeth" — human rights obligations over ethics principles. Marda's principal analytical contribution to the AI governance debate is the argument, most fully made in the ARTICLE 19 "Governance with Teeth" report (2019) and developed in the "Leap of FATE" FAT* paper (2020), that the shift from voluntary AI ethics guidelines to enforceable human rights obligations is not a matter of degree but of kind: the entire accountability architecture changes when the framework carries legal weight. The framing names the failure mode of ethics-only approaches — their voluntary and self-assessed character means that the most powerful AI deployers face no binding constraint — and proposes that existing international human rights law already provides the enforcement infrastructure that AI governance needs, without requiring the invention of a new regulatory regime. The "governance with teeth" phrase has circulated beyond Marda's direct output through the civil society AI governance field as a shorthand for the same critique.

• "AI in the majority world" / ethnographic methods as discipline. Marda's methodological argument — most sharply made in the Nature Machine Intelligence article (2021, with Narayan) — is that AI research overwhelmingly studies systems in Global North institutional contexts and then generalizes findings to non-Western settings where the social, legal, and infrastructural conditions are structurally different. The ethnographic research program she developed — documented in the FAT* '20 Delhi policing paper, the 2018 Philosophical Transactions A article on AI policy in India, and the ARTICLE 19 Myanmar CCTV report (2022) — is the corpus's clearest instantiation of the argument that these limits require a different research methodology, not just a geographic correction. The framing positions AI research in non-Western contexts as a discipline in its own right, not merely an application of methods developed elsewhere.

• "AI bias is not just a data problem." The framing Marda develops in her Forbes India op-ed (August 2021) and across her research register names a structural misdiagnosis in the AI fairness field: treating AI bias as a data quality or data representation problem that can be resolved by cleaning or diversifying training data, rather than as a system design problem that reflects and reproduces social structures. The claim grounds in her Delhi policing research — where the predictive system's bias cannot be corrected by better data, because the data is a record of a biased process — and generalizes to other AI deployments where the training data is not the site of the structural problem.

Organisational vehicle

Marda's public output has run through three successive institutional vehicles. At ARTICLE 19, the London-based international human rights organization where she established and led the inaugural international AI programme, she co-authored the foundational 2018 report with Privacy International, produced the "Governance with Teeth" and "Emotional Entanglement" reports, and built the programme's research agenda, policy engagement, and global team capacity. Alongside and after ARTICLE 19, she co-founded REAL ML (Research, Engagement, Advocacy & Lived Experience in Machine Learning) — a global non-profit whose name encodes its methodological commitment to lived experience alongside research, engagement, and advocacy — as the institutional home for the interdisciplinary practitioner community she was building across academic AI fairness research, civil society advocacy, and policy engagement. She currently serves as Director of Ecosystem Building at The Collaborative for AI in the Public Interest (the AI Collaborative), where she leads work on building worker power across the AI supply chain and led the programmatic strategy for Current AI, the $400M+ public-private partnership for AI in the public interest. Across all three vehicles, her engagement with IEEE and ACM standardization processes, the UN Human Rights Council, the EU AI Act process, and the Bletchley AI Safety Summit has carried the same core argument — "governance with teeth" over voluntary ethics — into the technical standards, international human rights, and AI safety governance streams simultaneously.

Why this is a Voice entry

A Voice entry is created here, rather than additional structure on the Person entry, because Marda's public-facing output is itself the load-bearing object the corpus needs to track: the Delhi predictive policing FAT* paper (2020), which is the corpus's clearest artefact of the majority-world AI research register and the ethnographic methods argument; the Nature Machine Intelligence ethnographic methods article (2021), which carried that methodological argument into one of the field's most prominent journals; the Emotional Entanglement report (ARTICLE 19, 2021), the corpus's anchor on the China emotion recognition market and the research that generated The Guardian's and Reuters' international coverage; the "Governance with Teeth" report (ARTICLE 19, 2019) and the "Leap of FATE" FAT* paper (2020), which together constitute the corpus's clearest articulation of the human-rights-over-ethics-principles argument; and the op-ed and policy-engagement register that ran from Indian mainstream media through the UN Human Rights Council, EU AI Act process inputs, and the Bletchley civil society seat. The corpus's Indian voice anchors previously covered the feminist-data-governance / AI-constitutionalism register (voice-anita-gurumurthy) and the civil-liberties-litigation register (voice-apar-gupta); Marda's empirical-researcher-and-FAccT-community register, her majority-world documentation focus, and her ARTICLE 19 AI programme anchor are structurally distinct from both. The corpus lacked any voice on the ARTICLE 19 AI programme; any anchor in the FAccT/FAT* interdisciplinary research community connecting academic AI fairness work to civil society advocacy; and any researcher whose published output specifically documents how AI systems function in Indian and Chinese institutional contexts. This entry gives all three registers their first anchor. Affiliations, institutional history, and biographical detail are recorded on the linked Person entry per the corpus's Person/Voice split.