Could AI become more Biased than the society It learns from?
(Reflections on UNESCO’s Outlook Study on Artificial Intelligence and Gender in South Asia) Last week, I was invited to the launch of UNESCO’s Outlook Study on Artificial Intelligence and Gender in South Asia at the UNESCO Office in New Delhi. As we design, build, and train LLMs, we are baking in the biases already present in our society while creating the risk that the AI layer influencing our lives becomes even more biased than the society it learns from. What does that mean? The AI layer shaping our society could eventually become more biased than the humans it learns from if we fail to identify, test, and address these biases continuously throughout the AI lifecycle. To reduce risks such as embedded bias, hallucinations, and harmful outputs, governance, testing, and risk management must be applied throughout the model development lifecycle, not only after deployment. As Responsible AI practitioners, one question keeps returning: Do we have sufficient legal and governance guardrails when organizations build or deploy AI systems? Studies like this play an important role in strengthening accountability. They provide evidence that helps policymakers, researchers and organizations identify systemic risks and reduce harmful bias before it becomes embedded at scale. One of the biggest challenges in global AI governance is balancing speed, complexity and local context. Training data is often collected across countries and cultures without sufficient attention to gender, caste, language, cultural diversity, or equitable representation. AI is becoming part of everyday life. Before organizations procure and deploy AI systems, they should understand how the underlying models were trained, what governance processes were followed, how the systems were evaluated and how their recommendations could affect citizens’ access to public services. Ultimately, this is about more than model performance. It is about preserving public trust in AI. In this context, both the UNESCO Recommendation on the Ethics of AI and periodic studies such as this are essential. They help us understand how AI is evolving, who is participating in shaping it and where governance and inclusion still need attention. About the report This is UNESCO’s first regional assessment of women’s participation across the AI ecosystem in Bangladesh, Bhutan, India, Maldives, Nepal, and Sri Lanka. Its central finding is striking: South Asia is producing more educated women than ever before, yet very few are reaching positions where AI is designed, governed, funded, or led. Top six findings Download the full report here: https://tinyurl.com/5ydujdhd
Designing for Trust in BFSI and Healthcare
Trust is often treated as a communication problem – something to be solved with better messaging, sleek branding or AI chatbots. But in BFSI and Healthcare, that approach doesn’t hold. Trust isn’t claimed; it’s experienced. In these sectors, interactions aren’t casual. A claim, a diagnosis, or a payment carries heavy emotional weight. Users aren’t “exploring options” – they are trying to be certain. THE EROSION CYCLE Most systems are designed to complete a process, not to reduce doubt. When flows prioritize data capture over clarity, we trigger a dangerous chain reaction: Gaps in understanding → Users filling gaps themselves → Guessing → Trust eroding. Think about a typical insurance journey. Technically, nothing is broken. The forms work. The “Submit” button triggers a response. Yet, the user feels uneasy because: The system works, but the experience fails to reassure. THE ANATOMY OF ASSURANCE In high-stakes environments, users aren’t looking for “delight” – they are looking for assurance. Real trust is built on three pillars:1. Clarity: What is happening right now?2. Control: What can I actually do about it?3. Predictability: What exactly happens next? We often see organizations try to “signal” trust through badges and copy. But a clear claims timeline builds more trust than a “We Care” banner. A guided medical report builds more trust than a beautiful UI. Our role as designers isn’t to remove the complexity required by compliance, but to translate it into something actionable. In BFSI and healthcare, users are not looking for delight. They are looking for assurance. In insurance, they buy the hope of support when things go wrong.In healthcare, they trust the system to help them recover.Design cannot change the policy or the diagnosis. But it shapes how both are understood. THE BOTTOM LINE Trust is tested most when things go wrong-a delayed claim or an unexpected medical outcome. These are the moments where transparency and clear reasoning can actually strengthen a relationship rather than shatter it. Trust isn’t a layer you add at the end of a project. It is built (or lost) in the quiet, functional moments of a journey. How is your team designing for certainty in high-stress user journeys? This post has been taken from the LinkedIn post by Nazim Iqbal
Celebrating International Literacy Day !
As we celebrated International Literacy Day on September 8 – it is a time to reflect what literacy means in today’s world. Possibly what semantic language was to literacy two decades ago is now design language. This is especially significant in a multilingual and multicultural country like India. The essence of design as a literacy force in today’s world is rooted in its power to drive cross-cultural communication, creativity, and problem-solving. Design transforms how we engage with information, navigate technology, and connect with each other across diverse cultures and languages. Understanding Design Literacy: Design literacy refers to the ability to understand, interpret, and apply design principles effectively. It’s not just about aesthetics, but also about how design shapes user experience, communication, and problem-solving.
Rise of the Human-Centered Design
We are now at an inflection in the digital evolution, and for design to be really impactful, it has to step out from the realm of designers. Which brings up the question: what would designer do, if everybody is a designer.
Why financial & banking sectors need human touch more than others: Part 2
30 March 2022. Financial institutions should ensure digital touch points are designed as conversations – real and human. This will earn them trust, and also inspire pride in their employees.
Evolution of human digital needs
15 March 2022. The pandemic accelerated digital adoption. The need to be digitally astute and accessible now cuts across sectors. Unfortunately, the war threatens to undo the progress. It is a threat that could stall the humanising of digital with the conflict pushing a dehumanising narrative in multiple dimensions.
How human is your organisation
20 February 2022. Business performance and revenue are no longer the only parameters on which business is measured. What was earlier considered pure human attributes are now parameters on which organisations are judged.
Why financial & banking sectors need human touch more than others: Part I
3 February 2022. To achieve growth in the post-pandemic world, BFSI sector companies need to respond to customers’ emotional needs like hope and trust
What I learnt about CX at my Abbu’s (father’s) shop
8 November 2021. My father built an optical shop in the busiest business district of Calcutta. That was my first real-world design school.
The loneliness of digital
7 November 2021. Increasingly, digital transformation of business and growth in digital is happening at the expense of human interaction. For brands, loneliness is an opportunity.
