The disproportionately dominant role played by men in the development of this technology, which leads to gender stereotypes and biases, is concerning
The last few years have witnessed the evolution of Artificial Intelligence (AI) into a powerful tool that enables machines to think and act like humans. Considering the ability of this technology to transform economies, the AI revolution holds unparalleled significance for a developing country like India, which has the second-largest population in the world. While recognising the tremendous potential that this technology holds, it is also important to understand and question the processes involved in the making and functioning of AI. The disproportionately dominant role played by men in the development of this technology which leads to gender stereotypes and biases, is concerning.
The growth of AI technologies: India being the fastest-growing economy in the world has a significant stake in the development of AI. Recognising this potential, the Government had in 2018 instructed the NITI Aayog to establish a national AI programme. Thereafter, the establishment of a national AI stack, the proposal for a National Mission on AI and the launch of a global AI summit in 2020 by the Prime Minister, are all evidence of the Government’s commitment to creating a robust AI infrastructure in the country.
With machine learning (ML) finding its application across sectors including healthcare, digital finance and education, the opportunities for job seekers are also abundant in this field. The robust Information Technology (IT) ecosystem coupled with talented human resource has the ability to convert India into an AI hub very soon. However, like any other man-made technology, AI artifacts too reflect the challenge of being cluttered with existing sociological biases. Such concerns need a strategic and tactful response to obviate the said technology from reinforcing the existing stereotypes and discriminatory social norms.
Feminist challenges to the proliferation of AI in Indian Society: It has been 25 years since the adoption of the Beijing Declaration and Platform for Action, that envisioned a peaceful and equitable world for all women. A significant gender bias still exists in our society. For instance, as late as in February 2020 the Supreme Court had to remind the Government that it’s arguments for denying women a position of command in the Army were based on stereotypes. Moreover, as per a recent report of UNDP titled ‘Tackling Social Norms’, gender bias is not a male problem as the findings of the report reveal that almost 90 per cent of the people (both men and women) hold some kind of biases against women and people from other marginalised communities. The challenges of social bias and other forms of discrimination against women and members of the trans community pervade all spheres of life. While the existing digital divide and under-representation of women in STEM (science, technology, engineering and mathematics) continue to pose significant concerns, AI and automation are throwing newer challenges for attaining substantive equality in the age of the fourth industrial revolution.
With 78 per cent of AI professionals being men, it is certain that male experiences inform and dominate all AI algorithms with negligible representation from the other communities. This power imbalance has severe adverse implications for the other sexes like affecting their access to jobs and loans by automatically vetting out their applications.
Even companies leading the drive of automation are unintentionally contributing to the existing social biases getting embedded in the AI systems. As highlighted in a 2019 report by UNESCO, it is not a coincidence that virtual personal assistants like Alexa and Siri have female names and they come with a default female voice. These features reinforce the existing social realities where a majority of personal assistants in the public and private sector are women.
The way forward: Some of the solutions that concern an overall better uptake of AI among women and minorities for India lie in designing these systems inclusively. This may seem fairly obvious on the face of it, but effectively it means using more inclusive training data wherever possible. For instance, women who have never formally learnt to read or write can use their smartphones to communicate. Such digital literacy — and India’s increasing digital penetration — can open up a world of opportunities. Voice assistants inbuilt into mobile devices allow women to access a range of tools — however, manufacturers need to ensure that the training data used by the voice assistant is inclusive. Indian accents can vary incredibly based on region, according to the Eighth Schedule of the Indian Constitution, India has 22 regional languages. These can vary further — based on geography and community. If the voice assistant is unable to help a woman due to dialect — then there exists a missed opportunity to include those who are not literate, but willing and able to use their smartphones for a range of purposes.
The solution is straightforward but the challenge lies in its implementation — there has to be emphasis on ensuring that voice assistants can recognise a range of Indian dialects. This is not particular to women — and is a much broader issue of regional inclusivity — but it can benefit women more due to the gender divide in accessing formal education. Manufacturers can pre-empt the digital divide from worsening, in this manner — and such a solution benefits society at large. Conversational AI can increase the number of customers on an e-commerce platform, help with information and knowledge access and enhance internet experience overall. There are many women who are older and often have not used devices as early on in life as their male counterparts. This leads to a stereotype that mothers are often not as adept at using their mobile phones, and some part of it can be rooted in reality — however, that can be changed with leveraging the power of conversational AI.
Another important element to consider with AI and disruptive technology is the nature of job loss — which will be borne unequally by women across the world. Gender justice in the workforce is a concern of immense importance which must be addressed when it comes to the future of work. Women have the highest participation rates in the informal sector. This is reflected across major sectors such as manufacturing, agriculture and services, where employment of the female labour force is mainly informal. Their jobs are more routine and less abstract than their male counterparts. Research in developing countries shows patterns of job-loss due to digital automation affect women-dominated jobs more than male-dominated ones. This indicates that there is an urgency to skill women so that they can take on jobs which require handling of the AI-based tools.
A majority of the female labour force should be skilled to handle such jobs across sectors. Empowering women technologically will enable them to benefit from emerging technology such as 5G, blockchain, robotics, digital communications and so forth — instead of being the reason for an economic landslide.
COVID-19 has affected women unequally — with women being on the receiving end of job loss, domestic violence and increased household responsibilities. When the post-COVID recovery for the economy begins, India must remember to focus on these gendered issues in order to leverage the potential of its working women — across formal and informal sectors.
In the truest sense of intersectional inclusivity, it is important to remember that inclusivity does not include only women in the strict social sense of the term — but also sexual minorities. The LGBTQIA+ community in India isn’t well-represented on forms which collect data for larger AI/ML analyses. Many Government forms only offer the ‘M/F’ categories to indicate gender and occasionally, ‘other.’ This leaves out an entire category of people who might identify elsewhere on the gender spectrum — or non-binary people, who do not identify as M/F.
It is vital to work towards weeding out the bias which exists while hiring from the general pool of applicants — by anonymising the profiles of candidates in terms of removing gender identifiers, limiting the skill sets to only those that are required for the job (and not inadvertently including skills which might be typically masculine) and locating non-traditional applicants.
These are some of the social disparities which must be kept in mind while designing and deploying AI systems in India. AI/ML solutions are vast and far-reaching — with the potential to revolutionise sectors such as business and governance, but their impact can be divisive, sexist and exclusionary if not implemented correctly.
