AI Can Deepen Gender Biases at Workplaces if Unchecked

Q) Before discussing this ILO report, can you give us a broad view about the employment of women in India? Also, where do Indian women stand in terms of labour participation rate when compared to other countries?

I think India's female labour force participation, and I've been talking about it for years now, has always been quite low. I think it's probably one of the lowest worldwide, if not the absolute lowest. And that's always been a worrying factor. While I have seen some periodic labour force survey data which suggests that it might have gone up a little bit and is hovering around a third. But I think it's still very low and quite pathetic. So I think just to give you a colour on the sectoral concentration, agriculture continues to be one of the largest employers of women. But largely, most of these people are informally engaged. Almost 55-60 percent of the working women are in agriculture, but mostly informal or unpaid. You'll see quite a few of them in the manufacturing. Besides that, you will see them in services, which is like education teachers, health care nurses, ASHA workers, and of course, retail, hospitality, and not to forget the domestic work segment. In the white collar, the concentration that's emerging for women would be a little bit in IT services, in BPM, HR, finance, very much skewed still towards entry and mid-level jobs.

Q) Now coming to the report, it says that occupations dominated by women are almost twice as likely to be exposed to Gen-AI compared to male-dominated ones. The report says around 29 % of female-dominated occupations are exposed to Gen-AI against just 16 % of male-dominated occupations. Which are these occupations, and is this global situation the same in the case of India?

I think globally some of the roles which stand exposed right now definitely are clerical roles, administrative roles, data entry assistant, customer service, big one, BPO call centers, content and documentation role, basic accounting, HR ops. Of course, these are roles where men, women are exposed equally. However, given that some of these sectors employ more women, hence it affects women. From an India context, I think it's even more relevant given that India has about 1.5 to 2 million in BPO, BPM kind of jobs with a very high share of female workforce, which is almost like 35, 40 percent. Very large base in clinical support and routine service role. So honestly, from my perspective, one can conclude that the trend is valid, of course, and in some cases amplified a bit in India. So these roles are typically task-heavy, they're repetitive in nature, hence easiest for Gen AI augmentation.

Q) Can you tell me about the men dominated sectors, you know, which are going to be less affected by AI?

Least exposed sectors, whether it's globally or in India would be construction, manufacturing, structural roads, logistics, transportation, mining, heavy industry, at least in the foreseeable future. And largely, the reason for them being least effective is obvious. I think they require physical presence, manual dexterity, real world unpredictability, and hence these roles at this point seem less likely. From an Indian advancing perspective, I think these sectors are already heavily male dependent, like almost 80 per cent plus workforce are male. AI disruption risk is clearly one of the ways of, this is one of the ways to prove that AI disruption risk is unevenly gendered here.

Q) The report says that under-representation of women in STEM jobs or jobs which are related to science, technology, engineering and mathematics is one of the factors that is leading to this kind of a situation. In India, do you think women are under-represented in STEM jobs?

Interestingly, that's not the case. I think the education pipeline is strong. Almost 43 per cent of STEM graduates in India are women. And it's one of the highest globally, which is great news. Workforce reality drops sharply because only 25 per cent So that's another problem. So while 43 per cent women are STEM graduates, when it comes to their representation in workplaces that are still low, it's almost 25 - 30 per cent in tech, even lower in core engineering, AI and leadership roles. There is definitely a leakage which is happening and that leakage is happening at predominantly mid-career level where there is a drop off on account of marriage or childcare. India doesn't have clearly defined re-entry pathways. There are of course cultural and workplace biases, which are one of the reasons that this is happening. So interestingly, in India's case, we don't really have a pipeline problem. It has a retention and progression problem.

Q) The report also brings a very crucial observation. It says that the systems which are trained on biased and incomplete data will be disadvantageous to women in recruitment, pay decisions, credit scoring, and access to other services. So we already have a lot of gender biases at workplaces. Are we going to see gender biases deepening with AI? Do you think so?

Highly probable if unchecked. So evidence globally shows that hiring and algorithms are favouring male profiles. The fact is, GenAI is dependent on data and data historically has been biased towards males because we have more data points from the male perspective. That's one reason credit scoring is disadvantaging women with nonlinear careers. From an India risk factor perspective, informal work history, which essentially means poor data visibility for women, because informal work has gone unrepresented, there is really no account or record of it. Also, there are lower digital and financial footprints here with relation to women. So that's another one. So from an outcome risk perspective, I think AI could scale existing biases faster and invisibly until and unless somebody is paying attention to the nuancing. So AI itself is not inherently biased, it learns from bias systems and data.

Q) The report says that with the right policies, social dialogue and gender responsive design, we can avoid reinforcing such discrimination. For this, the involvement of governments, employers and workers in training AI is important. How far is this happening in India? Are we actually even aware of this threat?

The current reality is that awareness is very nascent and fragmented. There has been some progress, like from a government perspective, there is an India AI mission, there is digital public infrastructure to focus on. From a corporate perspective, there are some DEI policies, but AI plus gender lens is very weak. I don't think anyone has gotten to that yet. What's missed is a structured gender audit of AI systems.

From a government perspective, think of mandated algorithmic audits for bias like ESG compliance. You can incentivize women in AI skilling programs at scale. You can build gender tag data sets via public digital infrastructure. From an industry perspective, I would say that why not embed gender by design in AI systems, audit hiding, pay and promotion algorithms as often as you can.

Also create return-ship programs for women in tech. And just to say from an overall ecosystem perspective, why not include women workers in AI training data sets and testing. And multi-stakeholder dialogue, which is government, industry, and academia who actually pay attention to all of that's going on and the possible risks.

Q) So do we need some policies to address this across all the segments, like the government, the industry, the people who are there, the data sets, everything together? Do we need a kind of policy and a framework for this?

I'm not a big believer that you have to jump into policy. First and foremost, there has to be an acknowledgement. There needs to be awareness. And there on, there has to be a willingness, a clear intent. I am forever apprehensive of jumping onto the policy bandwagon. I don't know how many people are even thinking about this possible challenge. And one of the issues is that while we talk a lot about AI, very few people actually know what it is. How you want to leverage it, whether it's in organizations or in a larger scheme of governance or building the country as a whole. I think we're still at a very, very superficial stage. But yeah, mean, some good signs, but I'm not somebody who would love to right now suggest policy changes.

Q) So at this point of time what will you suggest, what should be done?

It's very simple. AI will not create gender inequality, but it can either correct it or compound it. So the choice India makes now will define whether AI becomes an equalizer or an amplifier of bias.