The Future is Equal: Strategies to close the gender gap in AI

Generative AI technology is still at its nascent stage. There is a huge demand for professionals in AI research, AI engineering, and AI Product managers, who can bridge the gap between technology and business objectives.

By Lipika Dey 

During a visit to Glasgow University, about a decade back, a female faculty member from the university confided in me that the western society was still not favourable to girls studying science. I was happy to share that it was not so in India! According to data published by Ministry of Education in 2022, India still leads the globe with women accounting for 42.7% of STEM graduates. However, women account for only 27% of the STEM workforce in India, against a global average of 29.2%. Further, only 7 – 10% of leadership roles in tech industry are held by women. In academia, only 16.6 % of STEM faculty across 100 top universities of the country are women. The leaky pipeline does not definitely augur well for India’s AI mission that promotes ethical AI practices. Team diversity is the foundational principle for building systems that are fair, trustable and accountable.

The Ethical AI challenges

Along with the euphoria, generative AI applications have also been accused of bias and misrepresentation. Systematic explorations by researchers of MIT Media Lab, revealed biases in face recognition systems, that worked fine for light-skinned men, but failed to recognize dark-skinned women. The error could be traced to underrepresented samples in the training and test data. These systems, if deployed commercially, could lead to denial of services for the underrepresented categories. Omissions like the above can be extremely damaging to the society. An AI system trained with biased data can amplify digital divide by echoing the bias. That men and women perceive the world, think and act differently is well established by behavioural, cognitive and brain science. Brining in cultural contexts to decision making is also important.

Be it data bias, algorithmic bias or human bias, practitioners unanimously agree that the only way to ensure fairness in AI systems is to nurture diversity at all stages of development. In her book Invisible Women, Caroline Perez had highlighted how flawed data-driven policies and decisions can be, when systems were built on datasets win which women were conspicuous by their absence. History cannot be changed. But appropriate steps can be taken to augment the data before the next system is built.  Exhaustive assessment from diverse user perspectives also cannot be ignored.

Women in STEM – India-specific challenges

From social discrimination to lack of coaching facilities, girls aspiring to be engineers face several entry level barriers. Those who still manage to cross these, are challenged by entry level barriers to work, isolation in workplaces and dual burden of home and work. As much of tech hiring happens through competitive events like hackathons or coding contests or through long-term internships at distant places, social discrimination compounded by safety concerns, many girls lose out even before starting. With fewer women as colleagues, many young women describe STEM as a “boys’ club”. Isolation at work leads to fewer socialisation opportunities. This soon leads to women suffering from “Impostor syndrome”.  Women lose confidence and consistently undervalue their own capabilities to excel. Additionally, though the affirmative action in STEM education and hiring have led to increased opportunities for women, however, without appropriate sensitisation of all stakeholders, they also face peer group hostility that further vitiates the work environment and impacts their career growth negatively.  Maintaining work-life balance becomes the final nail in the coffin.  Many opt to sacrifice career progression in demanding situations. The significant growth in women-led tech start-ups may be a direct fall-out of this, however it’s a blow to diversity in the technology sector. Organizations lose possible future leaders.

Mitigation Initiatives

The country is still grappling to find the right solutions to retain women in STEM fields. There are quite a few national programmes to encourage and support diversity in STEM starting at schools, continuing through college and beyond. The supernumerary scheme for IITs have yielded very positive results with a significant rise in number of girls. Association for Computing Machinery (ACM) and the Institute of Electrical and Electronics Engineers (IEEE) run programmes to promote equity through mentoring, tutoring and networking sessions. Hackathons and coding contests are organized to encourage the culture of coding. Various organizations offer reskilling opportunities for women who want to come back from professional hiatus or face relocation challenges. But in an era of rapidly changing technology, it takes humongous effort to get back and take control. Social infrastructures have to be boosted to relieve the burden of dual responsibility on women. Social sensitisation also cannot be ignored.

The AI-powered future

Generative AI technology is still at its nascent stage. There is a huge demand for professionals in AI research, AI engineering, and AI Product managers, who can bridge the gap between technology and business objectives. Engineering tasks centre around development and deployment of AI systems, focusing on design, implementation and optimization for performance and scalability. Data scientists are responsible for building data repositories and gathering insights from them. Ethics AI consultants provide guidance to organisations on creating ethical AI plans and ensuring that AI systems are trustworthy i.e. fair, transparent, and accountable. Addressing security risks and privacy concerns are of prime importance.  Globally, women researchers and policymakers are at the forefront of advocating for AI regulations that protect users from discrimination and privacy violations. India needs to speed up. Take the lead girls.

(Lipika Dey, Professor of Computer Science, Ashoka University.)

Views are personal, and do not represent the stance of this publication.