AI for Disabled Indians: Ensuring Inclusive Tech for UPSC & PCS

06 Oct AI for Disabled Indians: Ensuring Inclusive Tech for UPSC & PCS

✎ Disability-inclusive AI requires pre-deployment bias testing, consented disability representation in training datasets, and adherence to RPwD Act-mandated accessibility standards to prevent exclusionary outcomes in governance.

Subject Relevance — Where This Topic Fits

• GS Paper II — Governance, Constitution, Polity, Social Justice and International Relations  |  GS Paper III — Science and Technology, Inclusive Growth and Issues Arising from their Design
• Prelims: Rights of Persons with Disabilities Act, 2016, Accessibility standards under RPwD Rules, 2017, AI bias and disability representation, Supreme Court judgment in Rajive Raturi case (2024), AccessEval benchmark for disability-inclusive AI evaluation, NClude platform for disabled users, Digital accessibility compliance mechanisms
• Essay: The ethical imperative of inclusive technology: Balancing innovation with equity, Governance in the age of AI: Ensuring accessibility as a fundamental right

Quick Revision: Disability-inclusive AI requires pre-deployment bias testing, consented disability representation in training datasets, and adherence to RPwD Act-mandated accessibility standards to prevent exclusionary outcomes in governance.

Why is this in the news?

The deployment of Artificial Intelligence (AI) systems by the Government of India, including in domains such as public services, recruitment, and healthcare, raises critical questions about disability inclusivity. Recent studies and judicial observations highlight systemic biases in AI models against disabled individuals, underscoring the urgent need for disability-aware AI governance. The Supreme Court’s 2024 judgment in the Rajive Raturi case further emphasised the legal obligation to ensure accessibility in digital systems, including AI, yet compliance remains inadequate. This issue is particularly salient given India’s ambitious AI investment targets and the expanding role of AI in governance.

Background

• The Rights of Persons with Disabilities (RPwD) Act, 2016, mandates accessible digital services for disabled citizens, with compliance required since 2019.
• The Supreme Court, in its November 2024 judgment in the Rajive Raturi case, held that accessibility is a facet of the fundamental right to life and dignity under Article 21 of the Constitution, and directed the Centre to frame mandatory accessibility standards within three months.
• Nearly two years post-judgment, petitioners have returned to court citing negligible progress in implementation, with the Chief Commissioner for Persons with Disabilities penalising 155 establishments, including government ministries, for non-compliant digital platforms.
• India’s AI ecosystem is rapidly expanding, with the Centre targeting $200 billion in AI investments over the next two years, necessitating robust frameworks to prevent exclusionary outcomes.
• Research indicates inherent biases in AI models, including higher error rates, negative tone, and stereotyping when disability-related inputs are processed, as demonstrated by benchmarks like AccessEval.

What is Disability-Inclusive Artificial Intelligence?

• Disability-inclusive AI refers to the design, development, and deployment of AI systems that are accessible, equitable, and free from bias against persons with disabilities, ensuring their meaningful participation in digital and governance ecosystems.
• AI systems must be trained on datasets that include genuine, consented representations of disabled individuals across diverse disabilities, conditions, and contexts to avoid underrepresentation or misrepresentation.
• Disability bias in AI manifests as errors, negative sentiment, or stereotyping when processing disability-related inputs, as evidenced by studies such as AccessEval and analyses of models like CLIP.
• Inclusive AI governance requires pre-deployment testing for disability bias, adherence to accessibility standards (e.g., WCAG 2.1, ISO/IEC 40500), and continuous monitoring of AI systems in real-world use.
• The RPwD Act, 2016, and its rules provide the legal framework for disability inclusion in digital services, including AI, but enforcement mechanisms remain weak and inconsistently applied.
• Judicial precedents, such as the Rajive Raturi case, reinforce that accessibility is a constitutional obligation, not a voluntary guideline, and mandate the Centre to establish binding standards for digital accessibility.
• Public procurement policies for AI systems must incorporate disability impact assessments and mandate accessibility compliance as a prerequisite for government adoption.
• Capacity-building initiatives for policymakers, technologists, and disability rights advocates are essential to bridge the knowledge gap and ensure disability-inclusive AI development.

Key Features

Feature	Significance
Accessibility in AI systems	Ensures disabled citizens can independently access government services, healthcare, and employment opportunities without reliance on external assistance.
Disability representation in training datasets	Prevents algorithmic bias by including diverse disability perspectives, ensuring AI models respond accurately to disabled users’ needs.
Mandatory accessibility standards under RPWD Act, 2016	Provides legal framework for enforcing digital accessibility, though compliance remains inadequate.
Supreme Court’s Rajive Raturi judgment (2024)	Reinforces accessibility as a fundamental right and directs the Centre to frame binding standards for digital services.
Bias in AI models (AccessEval benchmark)	Highlights systemic discrimination in AI responses to disabled users, affecting career guidance, healthcare, and service delivery.
NClude platform user survey (2,462 respondents)	Demonstrates real-world impact of biased AI on disabled individuals’ daily digital interactions and opportunities.

Why it Matters

Legal and Constitutional

• The Rights of Persons with Disabilities (RPWD) Act, 2016, mandates accessible digital services, aligning with the Supreme Court’s interpretation of accessibility as part of the right to life and dignity under Article 21.
• The Rajive Raturi judgment (2024) clarifies that accessibility rules must be binding, not merely advisory, ensuring constitutional accountability for digital exclusion.

Technological and Ethical

• AI systems deployed by the government must undergo disability bias testing to prevent systemic discrimination against disabled citizens in public service delivery.
• Inclusive AI development requires consent-based, representative datasets to mitigate biases that disproportionately affect marginalised groups.

Economic and Social

• Enabling disabled citizens to access AI-driven services independently reduces dependency on caregivers or intermediaries, enhancing socio-economic participation.
• Failure to address AI bias risks deepening digital divides, excluding disabled individuals from emerging job markets, healthcare innovations, and governance platforms.

Governance and Policy

• The Centre’s push for $200 billion in AI investment necessitates integrating disability-inclusive design principles into national AI strategies to ensure equitable outcomes.
• Regulatory oversight mechanisms must be established to monitor compliance with accessibility standards in AI systems, as mandated by the RPWD Act.

Challenges

1. Systemic Non-Compliance with Accessibility Standards

• Despite the RPWD Act (2016) and Supreme Court directives, compliance with digital accessibility norms remains sporadic, particularly in government ministries and public sector undertakings.
• The Chief Commissioner for Persons with Disabilities has penalised 155 establishments, including government entities, for non-compliance, indicating systemic failure in enforcement.

UPSC Link: GS2: Governance, GS3: Inclusive growth

2. Algorithmic Bias in AI Models

• AI systems exhibit significant bias against disabled users, as evidenced by the AccessEval benchmark, where models became more error-prone and negative when disability-related queries were posed.
• Image recognition models like CLIP show 15% lower accuracy for photographs taken by blind or low-vision users, and underrepresentation of disability-related objects (e.g., white canes) in training datasets.

UPSC Link: GS3: Science & Technology

3. Lack of Disability Representation in AI Development

• Disabled communities are often excluded from the design and testing phases of AI systems, leading to solutions that do not address their specific needs.
• Surveys of platforms like NClude reveal that disabled users face biased responses in career guidance and daily interactions, reflecting a broader failure in inclusive AI development.

UPSC Link: GS2: Social Justice

4. Legal and Institutional Gaps in Enforcement

• The Rajive Raturi judgment (2024) directed the Centre to frame mandatory standards for digital accessibility, but implementation has been delayed by nearly two years.
• Existing penalties for non-compliance are insufficient to drive systemic change, highlighting the need for stronger institutional mechanisms.

UPSC Link: GS2: Constitution

5. Resource and Infrastructure Constraints

• The Centre’s ambitious AI investment ($200 billion over two years) risks prioritising technological advancement over inclusive design, unless disability-inclusive frameworks are integrated from the outset.
• Limited awareness among policymakers and developers about disability-inclusive AI design exacerbates the problem.

UPSC Link: GS3: Infrastructure

Challenges — UPSC Perspective

Issue	Concern
Non-compliance with RPWD Act	Government and private entities fail to implement accessible digital services despite legal mandates.
Algorithmic bias in AI models	AI systems produce discriminatory outputs for disabled users due to underrepresentation in training datasets.
Exclusion of disabled communities in AI development	Lack of participation in design and testing phases leads to solutions that do not address real-world needs.
Delayed implementation of Supreme Court directives	Mandatory accessibility standards remain unimplemented nearly two years after the Rajive Raturi judgment.
Inadequate enforcement mechanisms	Penalties for non-compliance are ineffective in driving systemic change in digital accessibility.
Resource prioritisation in AI investment	Focus on technological advancement may overshadow the need for disability-inclusive design in AI systems.

Way Forward

• Establish a **disability-inclusive AI task force** under the Ministry of Electronics and Information Technology (MeitY) to audit government-deployed AI systems for bias and accessibility compliance.
• Mandate **consent-based, representative datasets** for AI training, ensuring proportional representation of disabled individuals and their specific use cases.
• Enforce **binding accessibility standards** for all digital public infrastructure, with clear timelines and penalties for non-compliance, as directed by the Supreme Court.
• Integrate **disability-inclusive design principles** into the Centre’s $200 billion AI investment framework, ensuring equity is a core criterion for funding eligibility.
• Strengthen **institutional oversight** by empowering the Chief Commissioner for Persons with Disabilities to issue binding directives and conduct periodic audits of AI systems.
• Launch **capacity-building programmes** for AI developers and policymakers on disability rights, accessibility standards, and bias mitigation techniques.
• Develop **public-private partnerships** to co-create AI solutions with disabled communities, ensuring lived experiences inform system design.
• Introduce **transparency mandates** requiring AI systems to disclose their accessibility compliance status and bias testing results for public scrutiny.

UPSC Value Addition

Prelims Practice Questions

Q1. Consider the following statements regarding the Rights of Persons with Disabilities Act, 2016:
1. The Act mandates accessible digital services for disabled citizens.
2. The Act was enforced in 2019 for compliance.
3. The Act empowers the Chief Commissioner for Persons with Disabilities to penalise non-compliant establishments.
4. The Act explicitly addresses the regulation of AI systems for disability bias.

How many of the above statements are correct?

1. Only one
2. Only two
3. Only three
4. All

Answer: Only three — Statements 1, 2, and 3 are correct as per the Act and subsequent judicial interpretations. Statement 4 is incorrect as the Act does not explicitly address AI systems or their bias.

Q2. Assertion (A): The Supreme Court, in the Rajive Raturi judgment (2024), held that accessibility is a facet of the fundamental right to life and dignity under Article 21 of the Constitution.

Reason (R): The Court directed the Centre to frame mandatory accessibility standards within three months, as existing rules were deemed non-binding.

In the context of the above two statements, which of the following is correct?

1. Both A and R are true, and R is the correct explanation of A
2. Both A and R are true, but R is not the correct explanation of A
3. A is true, but R is false
4. A is false, but R is true

Answer: Both A and R are true, and R is the correct explanation of A — Both the Assertion and Reason are factually accurate. The Supreme Court’s judgment in Rajive Raturi (2024) indeed linked accessibility to Article 21 and directed the framing of mandatory standards, making R the correct explanation of A.

Q3. Match the following provisions with their respective legal instruments:

Column I
A. Mandates accessible digital services for disabled citizens
B. Empowers the Chief Commissioner to penalise non-compliant establishments
C. Recognises accessibility as a facet of fundamental rights
D. Governs the use of AI in public services

Column II
1. Rights of Persons with Disabilities Act, 2016
2. Rajive Raturi Judgment (Supreme Court, 2024)
3. Digital India Act (proposed)
4. Chief Commissioner for Persons with Disabilities (Powers and Functions)

1. A-1, B-4, C-2, D-3
2. A-2, B-3, C-1, D-4
3. A-3, B-1, C-4, D-2
4. A-4, B-2, C-3, D-1

Answer: A-1, B-4, C-2, D-3 — A matches with 1 (Rights of Persons with Disabilities Act, 2016 mandates accessible digital services). B matches with 4 (Chief Commissioner for Persons with Disabilities has penalising powers). C matches with 2 (Rajive Raturi Judgment recognises accessibility under Article 21). D matches with 3 (Digital India Act, if enacted, would govern AI in public services).

Mains Practice Question

✍ The deployment of Artificial Intelligence (AI) in governance and public services presents both transformative potential and systemic risks for persons with disabilities. Critically examine this proposition with reference to the Rights of Persons with Disabilities Act, 2016, judicial interventions, and the ethical imperatives of inclusive AI development. (15 Marks)

Approach: MODEL-ANSWER SKELETON:

1. **Transformative Potential of AI for Disabled Citizens**
– AI as an enabler: screen readers, document summarisation, image recognition, and navigation aids.
– Case example: AI-powered tools assisting visually impaired individuals in accessing government services.
– Reference: Supreme Court’s recognition of technology as a tool for dignity and independence (Rajive Raturi Judgment, 2024).

2. **Systemic Risks and Disability Bias in AI**
– **Bias in Training Data**: Underrepresentation of disability in datasets (e.g., CLIP model studies showing 15% lower accuracy for images by blind users; 17x lower representation of assistive devices like white canes).
– **Algorithmic Stereotyping**: AI responses to disability-related queries (e.g., chatbot responses to career aspirations of blind individuals).
– **Exclusionary Design**: AI systems often lack accessibility features, exacerbating the digital divide.

3. **Legal and Institutional Framework**
– **Rights of Persons with Disabilities Act, 2016**: Mandates accessible digital services (Section 46) and reasonable accommodation (Section 2(r)).
– **Judicial Interventions**: Rajive Raturi Judgment (2024) linking accessibility to Article 21 and directing mandatory standards; Chief Commissioner’s penalising powers for non-compliance.
– **Gaps**: Delayed implementation, lack of binding standards for AI systems, and limited enforcement.

4. **Ethical Imperatives for Inclusive AI**
– **Participatory Design**: Involvement of disabled communities in AI development (e.g., co-designing datasets and interfaces).
– **Transparency and Accountability**: Auditing AI systems for disability bias (e.g., AccessEval benchmark for language models).
– **Consent and Data Governance**: Ensuring informed consent for data collection and representation.

5. **Way Forward**
– **Policy Measures**: Enactment of the Digital India Act to regulate AI in public services; mandatory accessibility audits for AI systems.
– **Institutional Strengthening**: Empowering the Chief Commissioner for Persons with Disabilities to oversee AI compliance.
– **Public Awareness**: Sensitisation of developers, policymakers, and the judiciary on disability-inclusive AI.

Balance of Views:
– Optimistic view: AI as a tool for empowerment if designed inclusively.
– Cautious view: AI risks deepening exclusion if bias and accessibility gaps are unaddressed.

Source: The Hindu

---

Generated by AanyaAi for educational purpose.

---

Related guides on our sites

• Best PSIR optional coaching for upsc
• Best PSIR optional teacher for upsc
• Best teacher of PSIR optional for upsc
• Best PSIR optional coaching in delhi for UPSC