Pradeep Kumar Muthukamatchi is a principal cloud solution architect at Microsoft. Recognized as a LinkedIn Top Voice for cloud computing.
AI is transforming everything, from coding to graphic design. As a result, it feels like we’re living in a time with limitless innovation. We are in an AI-driven era, which is rapidly evolving and brings new aspects to our daily lives. However, it’s crucial to make these solutions accessible and universal for all, especially for the disabled community.
According to the World Health Organization (WHO), over 380 million working-age adults live with disabilities around the globe, and unemployment rates among them are reaching up to 80% in a few regions. Currently, AI solutions are used to improve the lives of people with disabilities; however, fairness issues for people with particular classes of disability are often overlooked in training and evaluation data used for AI development. This creates a scarcity in disability data and becomes a major barrier to building an inclusive and responsible AI. I believe the need of the hour is to address the disability data scarcity and outline strategies for a more inclusive AI future.
Lack Of Inclusive Intelligence
AI solutions have the potential to improve the lives of so many people, but inclusiveness in AI for disabled communities has received less attention up to this point. Data is the backbone of any AI solution, but most AI models are trained with existing datasets that lack representation of diverse groups.
Recently, Forbes highlighted that the Center for Democracy and Technology has warned in its latest report that the lack of high-quality disability data in AI and algorithmic decision-making tools poses a significant risk of perpetuating and exacerbating existing barriers for people with disabilities in various aspects of life. This issue becomes more complicated when people with particular classes of disability are ignored when data is collected as they may represent a relatively small proportion of the community. The current state will lead to performance issues with AI models in recognizing and responding to the disabled community.
Navigating The Hurdles
• To overcome these technical challenges, there is a critical need to perform a risk assessment of current AI solutions for people with disabilities. This will help to identify the gap and act as a starting point for future research and development.
• More inclusive datasets can be created for testing and benchmarking AI models, and clear regulations must be established to protect the privacy of the disabled community.
• Synthetic data can be generated by simulation to create inclusive datasets. This approach can fill the gap by generating data by users simulating disabilities. Simulated data might not be perfect, but it is still appealing to make inclusive AI solutions.
• AI solutions must be designed with inclusivity as the core and a better bias mitigation approach that ensures fairness for disabled communities. One way to mitigate this is by designing a multimodal architecture that combines several AI models for text generation, speech and vision to accomplish more inclusive AI solutions.
• Designing custom AI solutions for particular user groups by feeding highly focused disability datasets can help harness AI resolution. For example, Be My Eyes is working with our company to make personalized AI models more inclusive for the over 340 million people around the world who are blind or have low vision.
• Recently, the European Parliament has approved a comprehensive framework to constrain artificial intelligence’s perceived risks and threats with the AI Act. Compliance and policies must be framed to have regulations in place to make technology accessible and responsible. The next generation must be educated to promote inclusivity in their innovations.
Why Inclusive AI Matters For Business
Making AI inclusive and accessible provides tremendous benefits and allows your products and services to reach diverse groups of users. Furthermore, it helps to comply with ethical standards and improve customer satisfaction and loyalty.
For instance, Mastercard is building a real-time inclusive AI solution to provide personalized assistance to small businesses that cater to diverse entrepreneurial needs. Adopting inclusive and diverse solutions in AI fosters innovation and helps organizations strengthen their diversity, equity and inclusion (DEI) efforts. This is evident from Accenture’s AI-powered inclusive skills-matching solution to identify and promote diverse talent from unconventional backgrounds, resulting in a more skilled and diverse workforce.
Embracing The Future
AI solutions offer a promising future. To make it more accessible and responsible, we must practice inclusive AI as a moral responsibility. For instance, Apple is making accessibility an integral part of its products. Users with physical limitations can select “Accessibility Options” during enrollment. This setting doesn’t require the full range of head motion to capture different angles and is still secure to use but requires more consistency in how you look at your iPhone or iPad.
As we move forward, our mission should be not only to build inclusive AI but to harness the power of AI to build a world that is inherently more accessible to all.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
