From a false sense of safety to resilience under uncertainty

Abstract

Understanding and acting upon risk is notably challenging, and navigating complexity with understandings developed for stable environments may inadvertently build a false sense of safety. Neglecting the potential for non-linear change or "black swan" events - highly impactful but uncommon occurrences - may lead to naive optimisation under assumed stability, exposing systems to extreme risks. For instance, loss aversion is seen as a cognitive bias in stable environments, but it can be an evolutionarily advantageous heuristic when complete destruction is possible. This paper advocates for better accounting of non-linear change in decision-making by leveraging insights from complex systems and psychological sciences, which help to identify blindspots in conventional decision-making and to develop risk mitigation plans that are interpreted contextually. In particular, we propose a framework using attractor landscapes to visualize and interpret complex system dynamics. In this context, attractors are states toward which systems naturally evolve, while tipping points - critical thresholds between attractors - can lead to profound, unexpected changes impacting a system's resilience and well-being. We present four generic attractor landscape types that provide a novel lens for viewing risks and opportunities, and serve as decision-making contexts. The main practical contribution is clarifying when to emphasize particular strategies - optimisation, risk mitigation, exploration, or stabilization - within this framework. Context-appropriate decision making should enhance system resilience and mitigate extreme risks.

Keywords: attractor landscapes; behavior change; change processes; complex systems; myth of mass panic; non-linearity; safety; security.

Figure 1

Schematic strategies for damage control, illustrating linear and non-linear impacts of adverse events. Purple: The linear case assumes an outcome that is gradual and directly proportional to the event’s magnitude. Blue: In this case, the system anticipates failure and is therefore capable of coping and mitigating it. Yellow: This line indicates scenarios where a protective barrier holds up to a point, until giving away catastrophically. Red: This line depicts the previous situation, in the worst case, where layers of protection have been omitted. The hexagon on the event magnitude scale marks critical values yielding a “tipping point” (see Figure 2) in the behavior of the non-linear curves, toward a new attractor. Gray: A threshold for a “zone of no return.” Dashed lines with arrows indicate the possibility of recovery by turning toward the original state. This is possible in the blue and yellow case, whereas recovery after passing the ruin threshold – whose exact position may not be known – is not possible. Figure adapted with permission from Bar-Yam and Seguin (2010).

Figure 2

Representations of a space of possibilities, where each tile is a system state. Solid and dashed lines indicate alternative paths that can be explored, stemming from an initial condition marked by the red dot in the center of the space. (A) The simple case with only one attractor. After the system is perturbed, it tends to return to the original state. (B) A situation with two attractors, whose depth is indicated by darkness of color. They “trap” and hold the system if it enters the basin of attraction (darker blue), making the system less likely – in proportion to attractor depth – to leave after being captured. A tipping point from one attractor to the next occurs when the system residing in one of the attractors crosses the striped tiles. (C) Space with one attractor leading to a systemic ruin risk (orange tile); trajectories that land there are permanently halted. The dashed trajectory depicts a route to the “safe” attractor, but the ruin state nonetheless remains as a possibility. (D) A space of possibilities, as seen from the policy maker’s perspective: only the “adjacent possible” states can be observed. Gray tiles represent states that are unknown (see text for details). Figure exapted from Lewin (1951, p. 92).