In my mid-twenties, after returning home from work, the janitor of the building complex I lived in asked me to call the police immediately as my mother had just been run over and was in hospital. After some frantic searches around public hospitals, I found her and learned she was clinically fine, but the shock of this experience led me to visit days later the exact location where she was hit by a motorcycle early evening. From this informal ‘inspection’ and conversations with residents and business owners, I learned that road accidents involving pedestrians occurred frequently at this intersection; however, no structural measures had been taken to mitigate these known risks. This experience then left me wondering i) What if I could share the risks, and evidence, of this intersection with a large audience in an easy, safe, and trusted way? What if I could learn what had happened before and monitor the structural and nonstructural measures taken by local officials -- and residents -- to mitigate these risks? How about creating something like this SenseCityVity project?

Ten years later, a toddler in Queensland died after being neglected and abused. That piece of news ‘hit me’ very hard as I had just become a father the year before. Pondering on what I could do to help this vulnerable group, as well as caretakers, I put together a research proposal to develop a mobile application that would collect ‘perceived risk inputs’ and provide, as a ‘weighted probabilistic output’, a set of recommendations for prevention and early intervention. Unfortunately, I have never had a chance to pursue this project titled: ‘Reducing the Cost Error of False Risks with Artificial Intelligence’. The main takeaway from this unrealized project was leading my curiosity to the field -- and potentials -- of artificial intelligence in the context of Error Management Theory so that I could incorporate those learnings into my future research projects.

I wrote the research proposal on child maltreatment because around that same time I was looking into the inherent and residual risks associated with the 2010/2011 Queensland Floods for my doctoral thesis. During my initial investigations on the early signs of this evolving risk, I came across a post shared in a forum from a concerned citizen urging residents living downstream the Brisbane River to evacuate due to the imminent water release from the Wivenhoe Dam. I became particularly interested in this message as it was uploaded days before the release of the official warning by authorities. As the underlying themes of my dissertation were risk perception and decision making, I could not stop wondering whether the number of deaths and missing victims would have been different, despite hindsight bias and related ‘behavioral traps’, had this post reached the people it needed to. How could this online forum have ensured that such an individual is reputable and trustworthy? How could this individual increase the trustworthiness of the piece of information he/she was sharing online? How could at-risk communities be informed about an emerging threat by an outlier? How could such an outlier be ‘intelligently’ validated through an automated classification and ranking system? This set of questions led me to explore the possibilities that heuristics, or rules of thumb, and networks create for informing individual risk perception and shaping decision making under uncertainty, which I later included in my thesis and served for my assessment on the role of eyewitnesses in spurring collective action during the COVID-19 pandemic titled: ‘Powering Social Media: Simple Guide for the Most Vulnerable to Make Emergencies Visible’.

Towards the end of my doctorate, and due to my growing interest in finding ways to efficiently collect and ‘intelligently’ rectify and validate individual risk perception for decision making, I audited ‘statistics’, ‘machine learning’, ‘algorithms’, and ‘artificial intelligence’ courses at the University of Queensland (UQ). Around that time, I was looking for a suitable location to test the hypotheses and limitations of the ideas I had in mind. After consulting with machine learning scholars in the U.S., I thought about designing a supervised machine learning application to help border agents detect high-risk travelers (note: I clearly understand such a project falls into the type of ‘social profiling’ and can further heighten social inequalities and indiscriminately target the most vulnerable, which is the reason why the European Union and many countries banned this type of artificial intelligence projects. However, my intent with this project is to further investigate not only the potentials but also the limits and risks of breakthrough discoveries and disruptive technology and how, I believe, they intersect with my specific research interests. The title of this proposal is ‘Detecting High Risk Threats and Improving Border Cross Experience with Machine Learning’.)

After the submission of my dissertation, UQ awarded me further scholarship to stay in Australia for another year, which made me the first UQ graduate student to be awarded with the Career Development Scholarship, so that I could continue pursuing alternative avenues to build bridges between the market and academia. I then joined UQ Ventures and, together with a doctoral candidate in the School of Engineering, co-founded a startup to help online marketplaces classify and rank reviews using Natural Language Processing to offer a better experience to their users. This enterprise attracted the attention of many engineering students from different programs across Queensland as well as CSIRO, which selected us as the first UQ startup to join their ON Prime accelerator program. CSIRO awarded us further monetary support after our team excelled in incorporating the design thinking principles, they had equipped us with, into testing and validating our ‘Minimum Viable Product’, or simple prototype. In the meantime, I had the privilege to meet many key stakeholders in the startup community in Brisbane and Sydney, including angel investors, venture capitalists, and decision makers like the then NSW’s Minister for Innovation, Science and Technology, the honorary Member of Parliament Matthew Kean. Unfortunately, this endeavor came to an abrupt ending when my co-founder and I realized that the timeline for funding this enterprise was not aligned with our family financial obligations once our scholarships expired. Nevertheless, the lessons from this transformational experience taught, and instilled in, me valuable soft skills, which I have since then applied in developing my project-, inquiry-based classes, as well as the way I approach and collaborate with my colleagues.

Finally, during the summer break of the Olympics in Tokyo, I spent most of my time sitting in a coffee shop trying to figure out whether there were any patterns that would help me explain and understand some of the most relevant and global political events occurring at that time. During this inquiry, I came across the work of Francis Heylighen and the provocative, and somewhat radical, book ‘Binding Chaos’ of Heather Marsh. Many of their research and propositions were quite original but it was the concept of ‘stigmergy’, which I had never heard of, that caught my attention. I then decided to conduct my own research on what stigmergy entailed and whether it would help me understand and, at least, explain some of the complexities I was witnessing around the globe. The more I explored the fundamentals of this ‘indirect coordination by autonomous agents in a mediated environment’, the more it helped me evaluate some of the contemporary social and political phenomena around me from a different and novel perspective. I have then started questioning whether it would be possible to replicate the autonomous and exploratory acts of ‘ants and termites’ on building and maintaining highly complex evolving systems by quickly responding and recovering from unexpected and massive environmental disturbances. This is when my epistemological -- and ontological -- interest for ‘human stigmergy’ was born and I became increasingly determined to explore its possibilities -- and limitations -- in explaining, predicting, and transforming our risk societies into resilient ones. The result of this inquiry led me to design the fundamentals of the ‘Stigmergy Network Theory’.

keywords: how AI improves risk perception and decision making; stigmergy network theory in simple terms; early warning systems using collective intelligence; machine learning for detecting real-world risks; human-centered AI for crisis response and resilience.