AI and Disability Rights: Enhancing Accessibility or Perpetuating Bias?

I. Introduction

As we move deeper into the era of artificial intelligence, the intersection of AI and disability rights is becoming increasingly significant. Technology has the potential to transform lives, offering unprecedented opportunities for accessibility, independence, and empowerment. However, it also poses challenges that must be addressed to ensure that these advancements benefit everyone, particularly individuals with disabilities. Article 1 of UDHR and Article 26 of ICCPR affirms that all individuals, including those with disabilities, are entitled to equality, dignity and rights protecting against discrimination and guarantees equal protection under the law, relevant to AI policies that may impact individuals with disabilities.

This exploration of AI and disability rights focuses on the promise of innovative solutions, such as smart assistive technologies and personalized support systems, while also examining the risks of marginalization and bias. As we navigate this evolving landscape, it is crucial to prioritize inclusivity, advocating for ethical AI development that respects the dignity and rights of all individuals.

II. Role of ai in disability rights

(A) Enhancing Accessibility

Artificial intelligence (AI) has the potential to significantly impact disability rights by enhancing accessibility, promoting independence, and fostering inclusion. Article 9 of the Convention on the Rights of Persons with Disabilities specify about accessibility which mandates that state parties ensure access to information, communication, and technology, which directly relates to the development of accessible AI. AI could provide autonomy and agency to a disabled individual. The basic idea behind accessibility is to let people free in a society that isn't built to suit their requirements².

More instances of how automation can increase accessibility to computers and productivity tools include the capacity to automatically create image descriptions and video captions. However, the development and implementation of AI technologies must prioritize inclusivity and accessibility.

• Speech Recognition: AI-powered speech recognition tools like Google Assistant, Amazon Alexa, and Apple Siri allow individuals with physical disabilities or mobility impairments to control devices using voice commands. These tools can also assist people with speech impairments when paired with speech recognition models designed to understand non-standard speech patterns. Other AI-powered tools like VoiceItt and Project Euphonia help individuals with speech impairments by training speech recognition models to understand atypical speech patterns.³
• Screen Readers and Text-to-Speech (TTS) Tools: AI enhances traditional screen readers, which assist visually impaired people in navigating websites, applications, and documents. AI-based systems improve natural language processing (NLP), making text-to-speech more accurate and responsive.⁴
• Real-Time Captioning: AI-driven captioning tools, like Google’s Live Caption or Microsoft’s real-time transcription services, automatically generate captions for audio and video content, making it accessible to deaf or hard-of-hearing individuals.
• Translation for Sign Language: AI-powered systems are being developed to translate spoken language into sign language, enabling communication between deaf individuals and those who do not know sign language. These systems can also help with video conferencing and live presentations.

(B) Education

AI can play a significant role in enhancing educational experiences for students with disabilities including accessibility in testing, content description and webpage interaction⁵. National policy for persons with disabilities, 2006⁶ aims to ensure the full and effective participation of persons with disabilities in all aspects of life, including access to assistive technologies and information systems powered by AI. AI can create customized educational experiences for students with disabilities, tailoring content and pacing to meet individual needs and learning styles.⁷

• Personalized Learning: AI systems can adapt learning materials to meet the needs of disabled students, offering customized content based on their learning pace and abilities. For example, students with dyslexia can benefit from AI-powered text-to-speech or font adjustments that improve readability.⁸ AI can analyse large data sets to identify trends and challenges faced by individuals with disabilities, informing policy decisions and advocacy efforts. Tools like the Disability Data Advocacy Toolkit help organizations collect and analyse data to support their advocacy efforts.⁹
• Accessible Online Courses: AI can help make online courses accessible to students with disabilities by automatically providing alternative text for images, generating transcripts for video content, and ensuring interactive features are compatible with screen readers.

(C) AI in Healthcare

AI is transforming healthcare by improving access to medical services for disabled individuals. For example, the development of mobility assistance systems is one of the most fascinating areas, where artificial intelligence plays a vital role for the creation of mind-controlled exoskeletons¹⁰. The "Mindwalker"¹¹ project from the Netherlands' Delft University of Technology is among the most prominent examples. With the help of artificial intelligence (AI), this research will create a mind-controlled exoskeleton that will restore independent walking to patients who have suffered spinal cord injuries. The exoskeleton gives those who were previously confined to a wheelchair new hope by translating ideas into physical motions by understanding the user's brain impulses. The "HAL" project¹² by the Japanese firm Cyberdyne is another example for AI’s exoskeleton project. People with disabilities are now able to execute physical tasks like walking, lifting objects, and ascending stairs thanks to this robotic exoskeleton that employs artificial intelligence (AI) to recognise electrical signals delivered from the brain to the muscles. People with impairments are living better lives because to this discovery since it gives them more freedom and movement. Others functions include

• Telemedicine and Virtual Assistants: AI-powered telemedicine platforms provide greater access to healthcare by enabling remote consultations. AI virtual assistants can also assist disabled individuals in scheduling appointments, refilling prescriptions, or navigating health-related information.
• Medical Image Analysis: AI systems can assist radiologists and doctors in analysing medical images, potentially identifying issues more quickly and accurately, which can be especially valuable for individuals who have difficulty traveling to multiple appointments.

(D) Employment and Remote Work

AI has the potential to make workplaces more inclusive and supportive of disabled individuals. Section 3 of the Rights of Persons with Disabilities Act, 2016¹³ prohibits discrimination against persons with disabilities in various domains, including education, employment, and access to goods and services, which encompasses AI technologies such as:

• Automation Tools: AI tools that automate repetitive tasks or assist with data analysis can empower disabled employees to work more efficiently. Voice-activated or adaptive technologies can also be integrated into remote work environments to ensure that people with mobility or visual impairments can perform tasks without physical barriers.¹⁴
• Job Matching Platforms: AI can analyse skills and match individuals with disabilities to suitable job opportunities, facilitating employment in an increasingly digital workforce like Mentra.¹⁵
• Workplace Accommodations: AI tools can help employers identify and implement necessary accommodations, ensuring a more inclusive work environment.

(E) Enhancing Communication

• Real-Time Translation: AI-driven translation services can bridge communication gaps for individuals with hearing impairments or those who speak different languages.
• Augmented Communication Devices: AI can power devices that assist individuals with speech difficulties, enabling them to communicate more effectively. For example, AI voice technology like Siri, Alexa, and Echo helps to facilitate communication with disabled persons. Persons with visual impairments can use these programs to describe text and visuals, and persons with brain injuries can communicate or understand things more readily thanks to text-to-speech (or vice versa) technologies. One such app is Google's Parrotron,¹⁶ an AI tool that helps people with speech impairments convert their stuttering speech patterns into natural speech.

III. How ai may pose risk for disabilities

While AI holds significant promise for improving accessibility for disabled persons, it also poses several risks if not implemented carefully. Both Article 2 of the UDHR¹⁷ and Article 2¹⁸ of the ICCPR ensures that every person can enjoy their rights and freedom without any discrimination. Disability is presented by a spectrum of characteristics, associated situations, stakeholders, interfaces, and technology, making it a challenging task to adequately serve.

A disabled individual may, for example, be devoid of limbs and exhibit altered posture and body type. It is possible for a blind individual to misinterpret visual information or cues. It's possible that people with hearing loss won't hear or follow spoken instructions or cautions. People with cognitive impairments could speak or behave differently, not be able to identify emotions, or have distinct speech or psychomotor patterns. It is well known that AI algorithms discriminate against people who have asymmetrical or uneven faces, unusual motions or gesticulations, speech difficulty, distinct communication styles, or assistive technology. The European Disability Forum¹⁹, the United Nations Special Rapporteur²⁰ on the rights of people with disabilities, and numerous other organisations and institutions expressed their concerns about these issues.

(A) Bias in AI Algorithm

Although algorithms do not produce biases themselves, they may reinforce societal injustices and cultural preconceptions. Bias in AI algorithms can significantly impact disabled individuals, leading to unfair treatment or exclusion.

Researchers examined biases against individuals with disabilities present in the natural language processing (NLP) algorithms and models they tested in their paper "Automated Ableism: An Exploration of Explicit Disability Biases in Artificial Intelligence as a Service (AIaaS) Sentiment and Toxicity Analysis Models."²¹ Systems, algorithms, and data sets are only a few of the development and deployment phases where bias might appear. There might be a number of factors, such as organizational structure and procedures, unconscious and intentional prejudice on the part of the development team, or lack of access to data for the target population. Thus, algorithms could bias against specific groups or provide predictions and outputs that are incorrect for some subsets of the population.

Algorithmic decision-making has the potential to worsen social inequities already present in other contexts.²² For instance, a person's history of prolonged unemployment may be partially due to disability discrimination; however, it may unintentionally affect the results of an algorithm used to screen job applications, worsening the individual's situation. If the training data lacks diverse representations of disabled individuals, the AI may not understand their needs or experiences, leading to inaccurate predictions or recommendations.

The design of algorithms can reflect the biases of their creators. If developers do not consider accessibility and inclusivity, the resulting algorithms may prioritize the needs of able-bodied individuals. An interdisciplinary approach that is guided by philosophy is justified by the combination of empirical and normative difficulties presented in a recent proposal by Altman²³ for assessing the fairness of algorithms employed in decision-making.

Addressing these biases requires inclusive data collection, user testing with diverse groups, and ongoing monitoring to ensure AI systems serve all users equitably. Collaboration with disabled communities in the design process is crucial for creating fairer AI solutions.

(B) Job Displacement and Economic Inequality

The right to work and protection against unemployment is guaranteed under Article 23 of UDHR, Article 6 of ICESCR, and Article 1(2) of the International Labour Organisation. But AI automation is projected to lead to job displacement in many sectors. Disabled individuals, who already face higher rates of unemployment, could be disproportionately affected by these changes. If AI systems are not designed to accommodate a diverse workforce, they could exacerbate existing economic inequalities for disabled people. Artificial intelligence is being used more and more in the workplace to automate worker monitoring and allow for constant worker surveillance. New data indicates that disabled employees may be red-flagged more frequently than people with no disabilities.²⁴

These AI technologies were intended to make the process of finding qualified candidates more efficient and less vulnerable to human bias in the recruitment process, where AI's use is growing. Evidence suggests that the data sets the machine is learning from are already "infected" with prejudice, though. Candidates that the organization has successfully recruited in the past have probably not come from historically underrepresented communities, such those with disabilities. As a result, the data sets lack them. This indicates that the computer will not consider them to be qualified applicants for the roles. Furthermore, a person utilizing a screen reader might not be able to use chatbots designed for interviews.

Unjustified rises in health insurance premiums can also result from flawed AI risk evaluations based on specific sorts of disabilities. For the majority of people, there are serious problems because of the transparency of the fundamental reasoning behind automated decision-making and machine learning. Furthermore, it is probably going to affect those with disabilities who are already marginalized in the markets for life and health insurance. Special Rapporteur Gerard Quinn²⁵ issued a warning regarding the seriousness of these risks and the potential for them to exceed the advantages if immediate action is not done.

Bias in automation makes this worse. A great deal of people mistakenly believe that computers and algorithms are flawless and capable of making logical decisions. As a result, there is an unwarranted degree of trust placed in algorithms, and instances of prejudice are often overlooked. For instance, worries about data protection can lead to the deletion or non-collection of data needed for decision-making. This may have the unfavourable effect of making it very difficult to decide whether, if discrimination against a disabled person has occurred.

(C) Data Privacy Concerns

Protection of privacy is guaranteed by Article 21 of the Indian constitution. In the case of K.S. Puttuswamy Vs UOI²⁶ where it is mentioned that the right to privacy a fundamental right safeguarded under Article 21 of Constitution of India. As of AI, it often requires access to personal data to function effectively. Article 17 of the ICCPR provides that no one shall be subjected to arbitrary or unlawful interference with his privacy, family, home or correspondence, nor to unlawful attacks on his honour and reputation. Everyone has the right to the protection of the law against such interference or attacks²⁷. For disabled individuals, this can mean sharing sensitive information about their health or disability status. Misuse, breaches, or poor handling of this data could expose individuals to privacy violations, discrimination, or targeted exploitation.

Many AI applications require personal data to function effectively. For individuals with disabilities, this could include sensitive health information, which may be collected without sufficient transparency. For instance, it was discovered in 2017 that a UK health trust had given Google's DeepMind access to 1.6 million patient records without the patients' express consent.²⁸

AI technologies, such as facial recognition and monitoring systems, can lead to increased surveillance, raising concerns about how data is used and who has access to it. There are some serious privacy concerns associated with the practice of using publicly accessible photos to build AI facial recognition databases. Large-scale face recognition datasets now in use have privacy consequences that should be brought to light via projects like Exposing.ai.²⁹ Numerous major corporations, such as Facebook and Microsoft, have ceased their facial recognition activities.³⁰ Privacy laws have been determined to have been broken by Clearview AI's use of a database of 10 billion photographs taken from the internet to provide face recognition technology for law enforcement reasons. These nations include Australia, Canada, France, and the UK.³¹

People with disabilities may face challenges in understanding complex consent forms or privacy policies, leading to potential exploitation of their data. Also, AI might make it easier to get personal information without adequate consent. Individuals with disabilities might be more vulnerable to data breaches, as their information could be misused or shared without their consent.³² For example, without the knowledge or agreement of up to 87 million Facebook users, Cambridge Analytica collected personal data between 2013 and 2018 for use in political advertising. "Smart" appliances, including refrigerators and cars, have the potential to sell user data to outside parties in addition to gathering it for performance enhancements. These devices could also put consumers at risk of hacker monitoring if they are not adequately secured.

(D) Erosion of Autonomy

AI can assist in daily tasks, but there's a risk that it might erode the autonomy of disabled persons if implemented improperly. For instance, smart homes or assistive devices powered by AI could make decisions for the user without their input or consent, reducing the person’s control over their environment. Another is relying on AI for daily tasks (like navigation or communication) can reduce a person's confidence and ability to act independently, potentially leading to a diminished sense of agency. AI-driven monitoring technologies in care facilities or homes can lead to a loss of privacy and autonomy, as individuals may feel constantly watched or controlled.

While designed to help, some AI-powered assistive devices may unintentionally restrict autonomy if they are overly prescriptive or lack customization options, forcing users into specific ways of interacting. If AI systems are not designed inclusively, they may not account for the unique needs of individuals with disabilities, leading to outcomes that do not align with their preferences or values.

Over-reliance on AI for communication or social interactions can limit opportunities for real human engagement, potentially leading to feelings of loneliness or isolation. Relying on AI tools can lead to increased data sharing and privacy risks, which can be particularly concerning for individuals with disabilities who may already be in vulnerable situations.³³

IV. Protection of rights of people with disabilities in the digital age

In the digital age, the protection of the rights of people with disabilities is critical to ensuring their inclusion and equal participation in an increasingly digital world. Several laws, policies, and innovations are geared toward protecting and advancing the rights of disabled persons in this context.

Indian law offers a range of exclusive rights and protections for disabled persons, primarily through the Rights of Persons with Disabilities (RPWD) Act, 2016. This legislation aims to empower and protect the rights of individuals with disabilities, ensuring their inclusion and equal participation in all aspects of life. National and State Commissions for Persons with Disabilities are responsible for monitoring and enforcing the provisions of the RPWD Act, 2016. The Chief Commissioner for Persons with Disabilities is tasked with addressing complaints and ensuring that the government complies with its obligations toward disabled persons. Countries and international bodies are enacting laws to address the unique challenges faced by disabled individuals in the digital space. United Nations Convention on the Rights of Persons with Disabilities (UNCRPD) emphasizes the importance of accessible information and communications technology (ICT) as a human right. Quinn in the Special Rapporteur on the rights of persons with disabilities³⁴ explained that states are required by law to "eliminate discrimination on the basis of disability by any person, organization, or private enterprise," as per the Convention on the Rights of Persons with Disabilities (ICRPD)³⁵. States are required by this duty to regulate the private sector's usage and development of artificial intelligence. Without going into a legal study of how the ICRPD's rules address such types of discrimination, he lists some instances of the discriminatory use of AI to people with disabilities.³⁶

By ensuring that disabled persons are part of the development process and by framing stronger regulations around data privacy, accessibility standards, and non-discriminatory AI usage can provide protection to persons with disability. Importantly, AI should enhance the human support system for disabled persons, not replace it entirely.³⁷ Currently underway efforts include wheelchairs that can avoid obstacles, software that speaks sign language, glasses that recognize objects, and canes that can tell when traffic lights are green.³⁸ People with disabilities can have better lives in so many ways thanks to AI. It may also result in the elimination of current work prospects and a requirement that everyone, including those with disabilities, retrain themselves in new skills or acquire new knowledge in order to succeed in a completely different sector of the economy, profession, or trade. By addressing these risks through thoughtful design, regulation, and ongoing collaboration with the disabled community, AI can better serve everyone.

V. Legal and ethical frameworks

In the context of AI and disability rights, several key legal provisions ensure the protection and empowerment of individuals with disabilities. The evolution of the legal rights of persons with disabilities in India has been shaped by a combination of legal reforms, social movements, and international influences.

The Indian Constitution, adopted in 1950, laid the groundwork for equality and non-discrimination through Articles 14 (Right to Equality), 15 (Prohibition of Discrimination), and 21 (Right to Life and Personal Liberty). These provisions have been interpreted to support the rights of persons with disabilities. Further, the Persons with Disabilities (Equal Opportunities, Protection of Rights and Full Participation) Act, 1995 which was a landmark legislation aimed at ensuring equal opportunities and protecting the rights of persons with disabilities. It focused on education, employment, and social security.

The National Policy for Persons with Disabilities, 2006 aimed to empower individuals with disabilities and promote their full participation in society. It emphasized inclusive education, skill development, and access to healthcare. India’s ratification of the UNCRPD³⁹ in 2007 has further strengthened the legal framework for disability rights, pushing for greater adherence to international standards and best practices. The Rights of Persons with Disabilities Act, 2016 replaced the 1995 Act and aligned Indian law with the United Nations Convention on the Rights of Persons with Disabilities (UNCRPD). It expanded the definition of disability, enhanced accessibility, and introduced new provisions for rights in education, employment, and social inclusion. Digital India Initiative promotes accessible and inclusive technology across the nation, focusing on digital infrastructure that considers the needs of persons with disabilities, ensuring AI is developed with inclusivity in mind.

The judiciary has played a significant role in upholding the rights of persons with disabilities through various landmark judgments. These legal provisions highlight India's commitment to protecting the rights of persons with disabilities, particularly in relation to the evolving landscape of AI. By ensuring compliance with these laws, stakeholders can foster an environment where AI serves as a tool for empowerment and inclusion.

VI. Key recommendation

The aforementioned study makes it abundantly evident that the development of AI will have serious implications for the enjoyment of rights by persons with disabilities. Accordingly, and in line with the guidance provided by the Human Rights Committee, CESCR and CRPD Committee, the State must make sure that these new technologies are designed and used in accordance with international human rights law in order to ensure equality and non-discrimination, respect for privacy, inclusion and accessibility, and the freedom to live independently for people with disabilities. If the human rights of the people with disabilities are to be placed at the centre of AI governance, the following practical actions are necessary.

• The State must ensure that the development of new AI technologies that are inclusive and accessible to persons with disabilities.
• The state should make sure how currently available AI technologies are being utilized to provide public services and to make sure that these technologies are properly modified to meet the special requirements of people with disabilities.
• States should not permit the use or sale of AI systems that pose significant hazards, and they should make sure that national rules pertaining to AI development and deployment include clauses prohibiting discrimination and safeguarding human rights. This necessitates the creation of public procurement norms that are inclusive of people with disabilities and human rights compliance, as well as the involvement of people with disabilities and their organizations in monitoring the effects of AI systems.
• Remind state of its legal responsibilities under international human rights law to effectively control the use of AI by public and private entities, with a focus on promoting equality and preventing discrimination, protecting private property, and facilitating accessibility, independent living, and community inclusion for people with disabilities.
• Increased awareness and advocacy by NGOs and disability rights activists to contributed to societal change and the recognition of the rights of persons with disabilities. This includes campaigns for inclusion and accessibility.

VII. Conclusion

The relationship between AI and disability rights presents both significant opportunities and challenges. While AI has the potential to greatly enhance accessibility, empowering individuals with disabilities to participate more fully in society, it also risks perpetuating existing biases if not developed thoughtfully. To truly harness the benefits of AI, it is crucial to prioritize inclusive design, actively involve disabled individuals in the development process, and implement robust safeguards against bias.

Currently, in India there is no specific law for regulating AI. Though, India committed to upholding the dignity and autonomy of persons with disabilities, promoting an inclusive society where they can lead independent, fulfilling lives in the era of Artificial Intelligence, now it’s time for inclusion of legal framework regarding Artificial Intelligence in the legal system. By fostering collaboration among technologists, policymakers, and advocacy groups, we can ensure that AI serves as a tool for empowerment rather than exclusion. Ultimately, the goal should be a future where technology enhances the lives of all individuals, regardless of ability, promoting equality and inclusivity in every facet of society.

*****