The Istanbul Innovation Days 2025 convened global leaders, thinkers, and innovators who are shaping institutional change and resilience in an increasingly complex world. This short series highlights key voices exploring the pressures driving institutional evolution and the barriers that stand in the way. From real-world examples to the challenge of anticipating crises before they unfold, these discussions explore how institutions can adapt to meet the demands of the future.
Today we’re meeting with Anthea Roberts, the founder and CEO of Dragonfly Thinking, an AI-powered analytics and decision-making support tool, which provides assistance with integrating foresight and understanding complex problems in more holistic ways.
According to Roberts, the challenge to overcome is not one specific issue, but the interconnectedness of risks across various domains, ranging from geopolitics to the environment and economics. Roberts argues for a need to move beyond the siloed thinking we are used to, both in terms of individual expertise and international understanding, such that we can tackle the complex and interconnected challenges of the 21st century.
A way to achieve this, says Roberts, is through what she calls ‘dragonfly thinking’. Based on the work of Philip Tetlock, dragonfly thinking urges us to analyse issues through a variety of lenses instead of the one or two that we are accustomed to. By looking at the world through only a single lens, we can be blindsided by other considerations, making it impossible to understand complex issues and predict forward. Hence, Roberts asks, “how can governments and international organizations, businesses and individuals, how can we all start to see through dragonfly eyes?”
According to Roberts, one way to achieve such multi-lens analysis is through leveraging AI tools, because they give us the ability to synthesize insights from across the world and across many different disciplines. By taking advantage of the power of AI to read and digest diverse and large amounts of content, individuals can break down silos between disciplines and regions, and integrate knowledge from a variety of fields and knowledge-bases to create new insights.
At the same time, most AI models are built using datasets lacking in diversity, still being dominated by English language and Western, US-centrism. As such, Roberts is interested in how we make the LLMs more representative in terms of culture and language. This will also allow for richer contextual understandings, which is important for intelligence analysis and strategic decision making, both in government and in business.
Finally, Roberts emphasises the importance of preparing for crises, as we cannot quantitatively predict the future. In her words, “if you want to be resilient and systemic, you don’t resile, you prepare.”
