AI for all: how technology can
drive social inclusion

Brazil’s CESAR innovation centre wants AI to create a more accessible world. Since 2019, its team has been working to develop a two-way video translator between Portuguese and Brazilian Sign Language, otherwise known as Libras. The project is of vital importance to Brazil’s 2.3mn-strong d/Deaf and hard of hearing community, who, CESAR hopes, will be able to use the service to access key services independently in their native language.

Creating a video interpreter is technically demanding. Linguistically, Libras is significantly different from Portuguese. Even the ability to capture signs via video can be knocked off course by variables including the user’s skin colour, clothing or lighting.

Two forces have been successful in overcoming these hurdles, says Vitor Casadei, Senior Technical Data Scientist Manager at CESAR. The first is the involvement of Brazil’s d/Deaf and hard of hearing community, who have spent thousands of hours sharing feedback and recording videos that can be used to train the translation system. “As soon as the community understood the project’s objective – and felt embraced by it – they became important project players,” says Casadei.

The project has also benefited from the arrival of generative AI. The translator has been in development since 2019, and recent AI innovation has been invaluable in helping the software navigate the fluidity of a living language. “This translation is only possible thanks to a unique artificial intelligence-powered solution,” Casadei says.

Building trust

CESAR’s project showcases the benefits that AI can bring to assistive technology – the services and products that support people with disabilities. Demand is growing rapidly, and better assistive tech will bring benefits across societies, says Ada Lopez, head of Lenovo’s Product Diversity Office. “It’s the technology that people need to share their thoughts, and we need those thoughts,” she says. “It's just good for humanity to allow more people to work, to think and to participate.”

While the project highlights the power of AI, it also spotlights potential pitfalls and challenges. Ultimately, AI is only as accurate or as intelligent as the data that has been fed to it, which means the translator relies on d/Deaf or hard of hearing people to share their knowledge and expertise.

However, there is no guarantee that communities will engage with AI. In 2023 a UK survey found just 19 per cent of people feeling optimistic about the impact of AI on society, while 35 per cent felt pessimistic. Meanwhile, high-profile cases of prejudiced algorithms – such as the AI-powered recruiting tool that systematically screened out female candidates – have also stoked concerns of inherent bias, especially for those who have experienced discrimination.

To fight this trend, Lopez and her team are encouraging developers to embrace feedback and set consistently high standards for emerging technology such as AI. They work closely with Lenovo’s Responsible AI committee to dig into data sets and algorithms and to examine ethical concerns. “It’s always progress over perfection,” Lopez says. “Sometimes we come up with the technology before there’s a standard. But we have to learn constantly, be able to take feedback and get things to a better place.”

Beating bias

If companies want to see AI reach its true potential, they will need to take more immediate steps to build trust and fight data bias. It will no longer be enough to address inclusion and accessibility issues after a product or service has already been made or developed: businesses need to embed these principles throughout their R&D timelines.

For many companies, the answer could be to invest in diversity in their own offices. Research has found that diverse companies do better. This is a trend that will only continue as businesses embrace next-generation tech, says Calvin Crosslin, Lenovo’s Chief Diversity Officer. Multiple viewpoints can give teams better perspectives on the data they’re gathering and how they use it. “If you have diverse teams, you have folks looking at things through a different lens: checking data sources and whether you have the right data on certain underrepresented communities,” he says. “We've caught mistakes early in the design phase of certain projects because of the diverse nature of our team. I don't think [diversity is] a hard sell.”

With the right approach to recruitment, companies can create a virtuous circle: more inclusion will allow businesses to create truly innovative products, while truly groundbreaking tech must be inclusive. These innovations will, in turn, help communities to gain new skills, access new opportunities and build new connections. “AI and technology can do wonderful things to bridge any digital divide among education and learning and healthcare,” says Crosslin. “Financially, it creates a whole new sector of jobs and employment. I believe we can have technology that levels the playing field.”