beyond policy lab 8
The Child You Cannot Algorithm: A Live Lab for Mothers and the Wisdom of Raising a Human Being
A group of mothers dedicated 90 minutes on a Saturday morning to reflect and offer wisdom from their journeys as mothers for the world of AI governance.
AG is the mother of a two-and-a-half-year-old girl. Akshita is from Tamil Nadu, grew up in Bangalore, and spent her high school years in Singapore and later years in the US, where her daughter was born. This brings her an interesting mix of philosophies into her journey as a mother. Raising a child taught her that she has a deep sense of intuition that is very loud. When she is able to hear and follow it, she finds that things are really aligned.
DS’s son is eleven years old. Her journey with her son shows her that he is probably bringing her up more than she, him. He has more questions than she has answers to. Raising her son taught her that power is not about control but about gently and safely guiding her son with the best of what she knows, to help him discern right from wrong, and to ease him into making decisions with wisdom.
Nanditha Ravindar’s son is nearly four years old. With ancestral roots in North India, Nanditha grew up in Chennai and speaks a dialectic Hindi at home while being deft at reading and writing in Tamil. Currently living in Hyderabad, Nanditha’s journey of motherhood combines a lot of influences. Raising a son has shown Nanditha that power hierarchies exist in all relationships - not just the workplace or classroom. She finds this to be a complicated question. Raising a son has also showed her that caregiving is both difficult and rewarding. One never gets to take a break from mothering or to put things off for later if they’re feeling unwell. It is also a gendered role, where most of the caregiving is done by women rather than men.
Pragnya Shabareesh is the mother of a three-year-old girl. Coming from Bangalore and having been raised in a conservative family, Pragnya’s own journey as a mother taught her the value of resilience. Navigating post-partum anxiety and depression and the resilience she received within, slowly and surely, endures as a powerful value she will carry for life.
Sindhu is from Chennai, with roots in southern Tamil Nadu. She is a single parent to a fifteen-year-old girl and a nine-year-old boy. Power, to her, is a layered and nuanced issue. Raising her children has required her to examine and unlearn inherited ideas, conditioning and systems around authority, obedience and power, which influenced how she was raised, and reconstruct her understandings and ideas around her own views and ideas. Sindhu’s experience of post-partum depression led her to grapple with a lot of questions internally. Social pressure often asks a lot of women, such as the expectation for mothers to simply “be happy” because generations of mothers before them supposedly endured the same journey without any sign of post-partum depression./
Tena Pick is from Croatia, her husband is from Chennai, but was born and raised in Dubai. They met in London and have spent the past eight years living in Bangalore. These global exposures bring in a lot of intercultural and intergenerational power struggles to the journey of parenting. This grounds Tena in the understanding that context is always queen for all lived experiences. A mother of a six-year-old boy, Tena learned about how having a strong-willed son who is every in his power, means that she also has to find where she is in her power in order to be able to navigate the relationship. Mothering has taught Tena about the invisibility of labor and the importance of making the invisible visible, and the fact that if one does not do that, nobody will do it for them. Not losing one’s identity in the process of mothering and giving care is a critical journey to keep up.
UN has a twenty-year-old daughter. Coming from a middle-class family, UN grew up thinking that parents have a lot of power, but becoming a parent turned that all around because she didn’t feel that way as a parent at all. She began to think that these little individuals have a lot more power than we expect to see in them, and that surprised her about parenting.
Naming and framing the problem
The group reflected on AI and all that it is becoming. With its capacity to generate, predict, surveil, decide, and replicate at scale, it shows both promise and peril. Key decisions around AI are being made by people in positions of power, and the tool continues to be shaped by them, while the dominant narratives around us go into training the tool. This is resulting in the erasure of multiple, complex realities, while the complicated tradeoff of fighting erasure is a visibility that attracts surveillance and control. AI is steadily being optimised to meet the needs of the powerful, while guardrails scaffolding it are slim at best.
As mothers, the group is confronted by a world order where AI is everywhere in their children’s lives. Education includes skills-building around AI, a necessity given that this technology has emerged and children cannot be left behind in a world where AI is embedded in their future. Recommendation algorithms decide what content a child is exposed to. AI tutoring systems are being deployed in schools and ed-tech platforms. Content moderation systems that are supposed to protect children fail frequently. Social media algorithms have been shown to amplify anxiety, comparison, and distress in pre-teens, voice assistants that answer a child's questions about the world before their parents have the chance to even hear the question. The group talked about how AI tools are part of children’s material realities without sufficient guardrails in place to give the children the emotional wisdom they need before they engage.
People making decisions about what AI does to children, whether in social media companies, ed-tech businesses, gaming, or the content platforms pre-teens inhabit are not accountable to parents in any meaningful way. They are accountable to shareholders. The frameworks being built to govern AI globally are largely written by people without children, or with children who have access to protections in the form of private schools, curated environments, informed parents with time and resources, active choices to opt-out of these technologies, that most children do not have. The governance gap between the children most affected and the people making the decisions is vast.
With this in place, the group reflected on three key questions:
You know your child in a way no algorithm ever will: Their particular fears, their specific way of being brave, the things that make them come alive. What does that knowledge tell you about what AI systems are getting wrong about children?
You make governance decisions about your child's life every day, about what they are exposed to, what they are protected from, what they are allowed to navigate on their own and when. What wisdom is in those decisions that the people building AI have never thought to ask you about?
Your child will live with the consequences of AI governance decisions being made right now for the next sixty or seventy years. The people making those decisions will mostly not. What would change if they did take this into account?
Creating a repository of wisdom
Each member in the group was invited to share one insight, practice, value, or hard-won understanding from their experience of mothering a child that carries a governance lesson. The themes that emerged were as follows:
Prioritizing what a child actually needs at each point in their lives to genuinely serve their becoming, rather than what they want or what is easiest to provide.
Being accountable to a child without surrendering the responsibility to protect their interests and safety
Surrounding a child with a village of actual and chosen family comprising people that respect and nurture the child's growth.
Drawing from accumulated experience, relationships, and practice over time to form value systems and develop character.
The practice of saying no slowly, with reasoning, as a counter to instant gratification and frictionless, always-accommodating systems.
Navigating the tension between community values and individualism modelled by global digital cultures that surround children, and the contextual wisdom that emerges from this journey.
The journey of reading a child’s silence and learning from experience that each particular quality of quiet tells you something different each time, and that silence is as communicative as verbal engagement.
Parents provide scaffolding/progressive autonomy to their children to learn, become competent, and then autonomous. Too much autonomy too early on can be maladaptive, and too little autonomy can be restrictive.
Serendipitous discovery gives the child the freedom to be their whole selves, instead of being confined to the assumptions a parent has about them.
Boundary practices that protect the child’s autonomy and agency while also giving them the wisdom to define what they want and don’t want.
Prioritising sensory experiences and play, rather than to normalise learning and skills-acquisition as the only goal of life.
Releasing power and desire for control and putting both in the hands of the child with the trust in both them and yourself as a parent that you are doing and being enough.
Offering information in ways that help them make an informed decision: This includes offering information on the upside and downside and making space for all the potentially unforeseen sides to something.
Space to come in with their opinions and ideas as they are without feeling the need to polish them, and then exploring what’s ethical, practical, and acceptable. It is important to surface ideas and scaffold them with wisdom without snuffing out the thrill of discovery.
Building a container for governance
AI is essentially driven by data, and we as humans produce this data through our actions and engagements with digital spaces. All these data points are collected and a line of best fit is drawn by the machine which learns patterns and produces outputs based on the internal architecture in the form of an algorithm. These patterns never acknowledge the diversities inherent within an individual and across a whole community. Mothering shows us that context is queen and that we are different people to different people. The same mother can raise each of her children in distinct and unique ways, and they may experience her mothering in unique and distinct ways. AI cannot understand rich context and its distinctiveness in shaping life and lived experiences, and it is essential for governance to recognise how the flat outputs of AI can be felt distinctly by different groups.
As AI is built on memory and the data accumulated to create memory, it is often the case that people’s data once taken into the tool’s learning cache, cannot be retracted. One of the participants reflected on a point in their journey as a mother where blogging online elicited a question from her daughter on how her mother’s readers knew about her, but she didn’t know about them. What does it mean for an AI tool to build a memory of people, places, and realities, when those involved seldom get to participate or consent to the creation of such a version? What might it mean to be remembered for or as something one would least prefer to be associated with? This discussion prompted a powerful idea for governance, which is the right to be forgotten.
It takes a village to raise a child. Even as mothering is a critical part of that journey, a child’s exposure to a larger ethos constitutes their upbringing. This in itself highlights three insights for governance. One, that governance should call for shared responsibility across the AI lifecycle, rather than demand single-owner accountability. Just as a child is shaped by their parents, teachers, relatives, peers, and institutions, AI governance should also involve every actor across the lifecycle and each must bear a shared but undiluted responsibility. Two, governance should be an ecosystem. A single safeguard is not sufficient, and a single approach for safeguarding for a heterogenous community is inadequate at best. Mothers consistently strive to build well-rounded, wholesome experiences for their children by making active governance decisions on the information, people, places, and experiences their children access and are exposed to. Similarly, governance should be a whole-of-system approach, much like how different members of a village offer different forms of care. Three, different stages need different levels and forms of oversight and governance. Raising a child is a verb, and involves progressively changing responsibilities as the child develops. AI governance can be based on lifecycles, with stronger scrutiny during development and deployment, continuous monitoring during operation, and immediate incident response when things go wrong, with a consistent focus on reassessment as capabilities and uses change.
Mothering also involves a consistently shifting understanding of risk that a limited, regulatory framework cannot address adequately. For instance, the EU AI act classifies AI based on risk, and the application of AI in biometrics, critical infrastructure, education and vocational training, employer and worker management, law enforcement and migration, essential services, and administration of justice and democratic processes are all considered high risk. However, for a mother, a generative AI model offering a child a list of ways to self-harm or take their own lives or offers misleading sexuality education is also high risk. AI governance should adopt a contextually informed understanding of risk that makes room for different lived realities and the ways in which they inform experience.
Another powerful insight the group brought to fore is that trust is relational. We don't determine that a child is safe by inspecting the child alone. We pay attention to the child's interactions, the environment and people around the child, the information and resources the child has access to, and the ways in which the child's behaviour and choices are shaped in relation to their material realities. One of the participants shared an interesting analogy from her son, who likened AI to fire and anger, and said that how it is shaped and how it is used determines whether it is safe, unsafe, good, or bad. This calls on us to ask key questions: Who is shaping it? Who is building it? Who is funding it? What incentives surround its use? What safeguards exist? What happens when these safeguards fail? Governance can learn a powerful thing from this, in that the safety of an AI tool depends on the socio-technical system around a model.
The mothers also talked about their journeys of mothering as involving some measure of unlearning, relearning, and reflecting, as well as a product of many influences, both social and cultural. They also recognised that they come into this journey with nothing to prepare them for it, because there is seldom a precedent in one’s lived experience that scaffolds their journeys as mothers. And yet, they learn from their immediate worlds, their own parents, and their reflections on realities within and beyond their immediate circles. This offers an important lesson for governance, which is to recognise, centre, and prioritise local knowledge as opposed to imposing a singular, “mainstream” form of knowledge as the only accepted threshold of knowledge to train, use, or govern AI.