Unsecurities Lab participant report : 31st March at Goldsmiths University

Jones, Nathan (2026) Unsecurities Lab participant report : 31st March at Goldsmiths University. [Report]

Text (usl_full_report-2_amended_final-findings)
usl_full_report-2_amended_final-findings.pdf - Published Version
Available under License Creative Commons Attribution.
Download (3MB)

Abstract

Executive Summary of Report A unique workshop where government practitioners and academics used immersive art to simulate decision-making under high levels of uncertainty. By observing a film featuring AI-generated biology, participants explored how diverse professional backgrounds—from evolutionary biology to crisis response—interpret ambiguous threats differently. • The exercise revealed that instinctive emotional reactions serve as vital analytical signals and that rushing toward a forced consensus can actually increase operational risk. • Participants also studied non-human intelligences, to model more resilient systems capable of surviving the volatile environments of the Anthropocene. • Ultimately, the findings suggest that cross-disciplinary collaboration and slowing downinterpretation are essential for managing complex technological and environmental hazards. • This experimental approach aims to strengthen security thinking by addressing the vulnerabilities inherent in AI-mediated evidence and shifting timescales. The workshop is most relevant to policy and security teams working with uncertain, AI-mediated or cross-domain evidence, where premature classification, weak signal loss, or over-fast consensus can create operational risk. The clearest entry point to this report is the Interaction Map in Output 1. It shows how the workshop moved from three different table readings of the same film to a shared diagnosis of the problem. Participants first responded through unease, alarm, empathy, anger and suspicion. These reactions were then tested through group discussion and developed into common observations about intervention, misreading, timescale, self-reinforcing evidence and classification failure. The main finding is that the workshop showed how groups can begin to handle uncertain evidence better by collaborating across disciplines. In Session 1, participants slowed down, questioned the material, compared different professional perspectives, and avoided settling too quickly on one explanation. These are practical skills for working with ambiguous evidence. In Session 2, the group used speculative models of resilience to imagine what systems will respond better to volatile and uncertain conditions. The discussion produced four useful capacities: remembering what has happened over long periods, coordinating across different perspectives, recognising when conditions have changed, and adapting when existing ways of operating no longer work. The transcript analysis gives the evidence for these findings. It shows how professional background shaped what participants noticed first; how each table built its account; where people changed their minds; how instinctive reactions became useful analytical signals; and which observations emerged independently across all three tables. Output 2 translates the workshop findings into five stress points for security thinking: unreliable evidence, mismatch between intervention and system timescales, systems feeding on their own outputs, the cost of forcing consensus, and the risk that what cannot be named cannot be protected. The final section explains how Security Lancaster can develop this approach further with policy, security and research partners, including bespoke workshops on issues with complex environmental, agential AI and cyber-physical overlaps.