Predicting tipping points – more harm than good?

This thought piece was inspired by a series of academic papers in which the authors predict tipping points within a system(1,2,3). They start ‘simple’ and then in subsequent papers make the challenge harder, and more realistic.

Tipping points – a quick refresher

Tipping points are thresholds that transition a system from its current state to a likely irreversible new state. The climate system is among the most susceptible to tipping points due to its highly interacting parts (complexity) and an ever-increasing human activity exerting pressure on it. The Thinking Ahead Institute has repeatedly focused on this topic in previous works4 remarking how delicate, complex and uncertain our system is.

Our inability to predict the future path of complex, non-linear systems leaves us with great uncertainty. One option is to ‘retreat’ to the present and reconsider our influence on the system.

Machine learning is a new superpower

The advent and the very recent surge of machine learning has proved to be a powerful tool across numerous problems whether recognising pictures of cats, human faces, or beating humans at the game of Go (unthinkable until relatively recently). It has also been applied to predicting tipping points. The authors of the papers noted above stated that “machine learning has great promise as a new and highly effective approach to accomplishing data driven prediction of […] both continuous gradual climate evolution as well as relatively sudden climate changes (regime transitions).”

The list of alleged benefits is long

The potential to provide decision-useful climate-related information on the back of tipping point predictions is enormous. If it were truly possible to know where and when tipping points were likely to occur, and how the system might behave beyond the tipping point, then we could focus on targeted interventions in specific regions of the planet. For example, knowing the regions where critical shifts are likely to happen can guide initiatives, implement climate adaptation measures, and protect vulnerable ecosystems.

These initiatives could include:

  • Water resource management strategies to counteract predicted drought or water scarcity in regions of the planet (eg water recycling, infrastructure upgrades, etc)
  • Coastal resilience measures to counteract predicted sea-level rise (eg planting mangroves, building or reinforcing sea walls)
  • Biodiversity conservations and regeneration to counterbalance predicted degradation of certain ecosystems (eg reducing local stressors like pollution, overfishing, reef resilience, marine protection, etc)
  • Disease prevention against predicted tipping point in disease transmission due to climate change (eg prioritisation of vaccination campaigns, targeted surveillance systems, etc)

These examples illustrate how knowing the specifics of where and when tipping points might occur could inform targeted policies and actions. All the examples are sensible actions to take in the face of climate change, but by tailoring interventions to specific regions or systems, the potential for effective mitigation and adaptation would increase, leading to more resilient and sustainable outcomes. This is, of course, true in theory. We do not know, as yet, whether predicting tipping points at this level of granularity will be possible in practice.

A panacea for all evils?

While it might seem obvious that being able to predict tipping points was a good thing, this is not necessarily true. Perversely, the knowledge could lead to:

  • Capital flight | if it is known that a country is heavily reliant on a particular industry, resource or export, then it is both exposed to dire consequences from a tipping point, and to investors withdrawing capital on any prediction that the tipping point could be near
  • Supply/demand imbalances and price volatility | Knowledge of tipping points may lead to conflicts over resource allocation and priorities. Different regions or interest groups expected to be disproportionally affected by tipping points may compete for limited resources, hindering cooperation and effective collective action
  • Social unrest | Awareness of impending tipping points can cause social unrest, as communities may experience fear, anxiety, and insecurity about the future. This can manifest in protests, conflicts, resistance to change and, in the extreme, to mass migration
  • False sense of security | In contrast to the previous point, the prediction of tipping points could sometimes create a false sense of security and complacency. If individuals and policymakers believe that they have ample time to address the issue before a tipping point occurs, they may delay taking necessary actions or implementing effective policies
  • Geoengineering | While geoengineering aims to counteract the extreme effects of tipping points, it could have unintended consequences and ethical implications (derived from their novelty and lack of empirical evidence/testing), leading to potential environmental and social risks

In short, we are not prepared for predicting tipping points

This thought piece was inspired by the idea that it is possible to predict tipping points. The major benefit, if we could do this in practice, would be the ability to apply our limited resources to the best mitigation and adaptation options. The two biggest negatives are, first, that reality is vastly more complex than any model and so there is no guarantee that we could predict real-world tipping points. Second, a world in which we could predict tipping points would require even more collective action, global alignment and cooperation; in short, more systems leadership. All elements we are significantly lacking.

We are not prepared to cope with such a theoretical and technological breakthrough because we still profoundly lack a comprehensive understanding of what tipping points entail, let alone the commitment to act to effectively address the associated risks and consequences.


[1] https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.120.024102 or you can refer to a more divulgation piece here

[2] https://pubs.aip.org/aip/cha/article/31/3/033149/342213/Using-machine-learning-to-predict-statistical

[3] https://pubs.aip.org/aip/cha/article-abstract/33/2/021101/2875964/Predicting-climatic-tipping-points

[4] Here and here