How To Design an Ethical AI Chatbot
The Equity & AI Learning Series is a six-session collaborative learning initiative led by Cooperation Canada, in partnership with Salanga, created in response to the rapid adoption of AI across the non-profit and international cooperation sector.
How do we design ethical AI chatbots that are grounded in community realities and safe, useful and meaningful in practice?
As organizations across international development increasingly explore the use of AI chatbots, one important question remains: What does it actually take to make these tools safe, ethical, relevant and useful in real-world contexts?
The fifth session of the Equity & AI Learning Series explored this question through the experience of Girl Effect, a global nonprofit organization, designing media and digital tools, including AI chatbots, to support adolescent girls across Africa and South Asia.
Led by Karina Rios Michel, Chief Creative & Technology Officer at Girl Effect, the session offered a practical look at the organization’s journey from early experiments to deploying AI chatbots at scale. It highlighted the focus, time and intentionality required to develop systems that are grounded in the realities of the communities they are designed to serve.
Girl Effect’s chatbots evolved from more than a decade of listening to the questions girls and young people were asking through advice columns, radio, SMS and digital platforms and the questions about relationships, health, identity and everyday challenges formed the foundation for what the chatbots eventually became. The session also explored why AI should not be treated as a standalone intervention, as participants examined the importance of building human support into AI systems from the beginning. The session also highlighted the importance of continuous, deeply contextual co-design, adapted to local languages, mixed-language use, tone, cultural norms and humour so responses feel relevant rather than scripted or external.
Moderation and safeguarding processes, partnerships with service providers, escalation pathways, vetted content and continuous testing, all surfaced as required to make AI safe and meaningful in practice with safety presented as a layered responsibility rather than a single feature. At the heart of the session was a recognition that ethical AI chatbots are not created simply by choosing the right technology, but built through listening, co-design, human support, safeguarding, local knowledge, strong partnerships and a willingness to learn and adapt over time.
Thinking about what it really takes to build AI tools that are safe, useful and responsive to the realities of the communities they are meant to serve? We have captured the session’s most important lessons, reflections and practical insights in the free learning material below:
