Classifying Global Catastrophic Risks

Abstract

We present a novel classification framework for severe global catastrophic risk scenarios. Extending beyond existing work that identifies individual risk scenarios, we propose analysing global catastrophic risks along three dimensions: the critical systems affected, global spread mechanisms, and prevention and mitigation failures. The classification highlights areas of convergence between risk scenarios, which supports prioritisation of particular research and of policy interventions. It also points to potential knowledge gaps regarding catastrophic risks, and provides an interdisciplinary structure for mapping and tracking the multitude of factors that could contribute to global catastrophic risks.

Full text

Accepted Manuscript
Title: Classifying Global Catastrophic Risks
Authors: Shahar Avin, Bonnie C. Wintle, Julius Weitzd¨orfer,
Se´
an S. ´
Oh´
Eigeartaigh, William J. Sutherland, Martin J. Rees
PII: S0016-3287(17)30195-7
DOI: https://doi.org/10.1016/j.futures.2018.02.001
Reference: JFTR 2277
To appear in:
Received date: 26-5-2017
Revised date: 26-1-2018
Accepted date: 12-2-2018
Please cite this article as: Shahar Avin, Bonnie C.Wintle, Julius Weitzd¨orfer, Se´
an S. ´
O
h´
Eigeartaigh, William J.Sutherland, Martin J.Rees, Classifying Global Catastrophic
Risks, Futures https://doi.org/10.1016/j.futures.2018.02.001
This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.

1
Classifying Global Catastrophic Risks
Shahar Avin1, Bonnie C. Wintle1, Julius Weitzdörfer1, Seán S. Ó hÉigeartaigh1,
William J. Sutherland2, Martin J. Rees1
1Centre for the Study of Existential Risk, University of Cambridge, UK.
e-mail: sa478@cam.ac.uk
2William J. Sutherland is professor of conservation biology in the Department of Zoology, University of
Cambridge, UK.
We present a novel classification framework for severe global catastrophic
risk scenarios. Extending beyond existing work that identifies individual
risk scenarios, we propose analysing global catastrophic risks along three
dimensions: the critical systems affected, global spread mechanisms, and
prevention and mitigation failures. The classification highlights areas of
convergence between risk scenarios, which supports prioritisation of
particular research and of policy interventions. It also points to potential
knowledge gaps regarding catastrophic risks, and provides an
interdisciplinary structure for mapping and tracking the multitude of
factors that could contribute to global catastrophic risks.
Highlights:
● Classifying global catastrophic risks according to critical system affected,
global spread mechanism, and prevention and mitigation failure, provides
a novel means of framing risks
● By concentrating on component factors that contribute to catastrophic
risks, the classification system highlights convergent risk factors that
merit prioritisation and also uncovers potential knowledge gaps.
● The classification system can structure an ongoing, dynamic process of
knowledge aggregation and horizon scanning with policy implications.
1 Introduction
In our uncertain times it is good to have something we can all agree on: global
catastrophes are undesirable. As our science advances we gain a better
understanding of a broad class of global catastrophic risk (GCR) scenarios that
could, in severe cases, take the lives of a significant portion of the human
population, and may leave survivors at enhanced risk by undermining global
resilience systems (Bostrom, 2002; Rees, 2003; Posner, 2004; Bostrom &
Ćirković, 2008; Tonn & MacGregor, 2009; Baum & Tonn, 2015). Much progress
has been made in identifying individual GCR scenarios, and in compiling lists of
the scenarios of greatest concern, but there is currently no known methodology
for compiling a comprehensive, interdisciplinary view of severe global
catastrophic risks. While a fully complete list of GCRs may remain beyond reach,
we present here a classification framework designed specifically to draw on as
ACCEPTED MANUSCRIPT

2
broad a knowledge base as possible, to highlight commonalities between risk
scenarios and identify gaps in our collective knowledge regarding global
catastrophic risks.
To date, research on global catastrophic risk scenarios has focused mainly on
tracing a causal pathway from a catastrophic event to global catastrophic loss of
life (Asimov, 1981; Bostrom & Ćirković, 2008; Coburn et al., 2014; Turchin, 2015;
Cotton-Barrat et al., 2016). Such research has been fruitful in identifying and
assessing a range of such GCR scenarios. Some severe GCR scenarios have posed
a persistent threat to humanity since our emergence as Homo sapiens (e.g.
impact by a 10km astronomical object, or a volcanic super-eruption of 1000 km3
of tephra). Other scenarios have increased in likelihood following human
population expansion and the accompanying increase in resource demands (e.g.
natural pandemics or ecosystem collapse). In addition, novel GCR scenarios can
accompany new technologies: some of these are relatively well established (e.g.
“nuclear winter” or an engineered pandemic); others are more speculative (e.g.
accidents in or weaponisation of advanced artificial intelligence, or
environmental shocks from ill-judged geoengineering efforts aimed at mitigating
climate change).
However, compiling a comprehensive list of plausible GCR scenarios requires
exploring the interplay between many interacting critical systems and threats,
beyond the narrow study of individual scenarios that are typically addressed by
single disciplines. The classification framework presented here breaks down the
analysis of GCR scenarios into three key components: (i) a critical system (or
systems) whose safety boundaries are breached by a potential threat, (ii) the
mechanisms by which this threat might spread globally and affect the majority of
the human population, and (iii) the manner in which we might fail to prevent or
mitigate both (i) and (ii). For example, a major astronomical impact may lead to a
global catastrophe if we lack the technology to deflect it (mitigation failure), and
it raises a cloud of dust that spreads around the world (global spread
mechanism), and that cloud of dust blocks sunlight for a sufficient length of time
to undermine the global food system in a manner that we cannot overcome
(critical system affected). Other scenarios will have different combinations of
one or more mitigation failures, one or more global spread mechanisms, and one
or more critical system breaches.
In order to gain a holistic picture of potential global catastrophes, knowledge
about each of the three system components needs to be explored and shared. By
first constructing a classification from the broad range of known critical systems,
global spread mechanisms, and prevention and mitigation failures, and then by
classifying known GCR scenarios according to these dimensions, we aim to: (i)
showcase the GCR relevance of a variety of scientific disciplines, (ii) highlight
how commonalities between threat scenarios have research and policy
implications, and (iii) highlight areas where there are potential gaps in our
knowledge of global catastrophic risks. We also propose concrete steps for
coordinating the broad-based, interdisciplinary research required to meet the
ACCEPTED MANUSCRIPT

3
challenges highlighted by the framework.
2 Critical systems
We define a “critical system” as any system or process that, if disturbed beyond a
certain limit or scale, could trigger a significant reduction in humanity’s ability to
survive in its current form (see Figure 1).
Building on the “life support systems” outlined in the research on so-called
planetary boundaries (Rockström et al., 2003; Steffen et al., 2015) (many of
which appear in our biogeochemical group), and their potential links to GCRs
(Baum & Handoh, 2014), we identify critical systems and processes that, if
disrupted, would affect human ability to survive. While we aim for
comprehensiveness and minimal overlap, we acknowledge that different
systems overlap. For example, while the processes affecting ocean acidity have
direct effects on ecosystem stability and thus human life, there is significant
overlap (causally, structurally and academically) with the global water cycle,
carbon cycle and sulphur cycle systems.
In our classification framework, critical systems are grouped at different levels in
a hierarchy, such that “higher-level” systems rely on the functioning of those at a
“lower-level”. Thus, the framework builds up from the stability of life-supporting
physical systems, through cellular and other systems, right up to species-wide
ecological and sociotechnological systems. “Lower-level” systems are directly
linked to human survival (which relies on functioning anatomical systems, which
in turn relies on cellular systems, etc.). “Higher-level” systems, especially
technology-enabled ones such as the food and health systems, help maintain the
human population at its current size, and provide resilience. If these “higher-
level” systems were to be disturbed significantly in some scenario, e.g. through a
severe and prolonged disruption to utilities networks (such as water and
electricity), or through shock effects (such as social unrest), these could cause
more harm than the system disturbance itself.
Identification of critical systems, and their cross-links, could also come from
historical and archaeological study of more limited instances of human
population collapse. For instance, the collapse of the Easter Island civilisation
shows how excessive resource extraction (of palms for the making of canoes) led
to ecological degradation, undermining primary production and food chains,
which in turn led to failure of the Easter Island society’s food system (Morrison,
2006). Further study of each critical system requires specialised expertise, often
in more than one domain, as there is no one-to-one mapping from scientific
disciplines to critical systems. Future work, conducted with collaboration with
the wider scientific community, could lead to the demarcation of safe operating
bounds for each critical system, following the example of Rockström et al.
(2003).
ACCEPTED MANUSCRIPT

4
Figure 1 Classification of Critical Systems aimed at identifying Global Catastrophic Risk scenarios.
Systems are grouped at different levels, arranged from “lower level” to “higher level” in a clockwise fashion
starting with the “Physical” group on the top-right.
3 Global spread mechanisms
For many critical systems, a failure of some instances of the system, e.g., regional
crop failure, would fall far short of posing a GCR. In severe GCR scenarios, the
failure of critical systems is coupled with some mechanism by which this failure
spreads globally, thus potentially threatening the majority of the human
population. In the framework, we separate the analysis of global spread
mechanisms from the analysis of critical systems (Figure 2). This separate focus
on global spread allows us to identify relevant mechanisms (and means to
manage or control them) as targets of study meriting further attention, and
highlights interesting commonalities.
A critical system failure can spread globally without human intervention: some
astronomical objects or events are sufficiently massive to have direct global
effect, while other threats can spread through the dynamic systems of the
natural environment, such as the air- and water-based dispersal systems. Dust
and toxins could be spread naturally even if they do not replicate, though of
course a self-replicating threat (e.g. a virus that affects multiple species of fish)
ACCEPTED MANUSCRIPT

5
could couple with a dynamic system (e.g. ocean currents) to achieve much faster
spread.
In addition to natural spread, many risk scenarios, and especially emergent risk
scenarios, rely on the highly connected nature of our species, both materially and
conceptually. A modern pandemic can spread through airports and other mass-
transit hubs of the globe-encompassing transit network, thus coupling a
biological replicator (this might be, e.g., a bacterium itself, or a biological vector,
e.g. a mosquito) to a highly connected anthropogenic network. A cyber attack can
cascade through global critical systems at the speed of digital communication,
shutting down health and security systems, and undermining resource extraction
and utilities by disrupting mines and power plants (a digital replicator, such as a
computer worm, could speed up the spread rate and reach).
Access to information can play a more abstract, but no less important, role in the
spread of critical system failure. The widespread, and growing, access of
individuals and groups across the globe to ideas, schematics, and manufacturing
capabilities (e.g. Do-It-Yourself, or DIY, biology) through digital and cultural
exchanges (e.g. online fora), enables novel hypothetical GCR scenarios. Such a
scenario could start with, say, the accidental or malicious release of a home-
grown pathogen, or the one-sided deployment of geoengineering efforts in an
attempt to mitigate climate change. Some ideas encourage their own spread, e.g.
schematics for communication devices, or ideas that encourage further sharing
of those ideas (e.g. ideologies or viral videos), coupling cultural replicators with
human interaction networks.
Figure 2 Classification of Global Spread Mechanisms relevant to Global Catastrophic Risk.
Table 1 illustrates how analysis of critical systems and analysis of global spread
mechanisms might be combined into a single classification framework. The table
presents a mapping from eight hypothetical GCR scenarios to the critical systems
that are most likely to be undermined in each scenario, for each type of global
spread mechanism. We have chosen a selection of severe GCR scenarios that are
(i) familiar, (ii) considered plausible, and (iii) cover both natural and
anthropogenic threats. This is far from a comprehensive list of scenarios, as the
very framework presented here aims to help explore possible scenarios.
ACCEPTED MANUSCRIPT

6
4 Prevention and mitigation failures
Analysing GCR scenarios along the dimensions of critical systems and spread
mechanisms draws significantly on our understanding of the natural world and
technical systems, and complements existing endeavours to classify risks of a
smaller scale (IRDR, 2014). Holistic risk management, however, must take into
account the human elements that moderate GCR through prevention and
mitigation efforts, and how these efforts might fail. The challenge of preventing
global catastrophes thus requires integration of the work and expertise in and
between the natural and the social sciences, on a global scale.
A particularly comprehensive existing risk management framework with such
integrative characteristics and international scope is the Sendai Framework for
Disaster Risk Reduction (SFDRR), adopted by 187 UN member states in 2015
(UNISDR, 2015). Although developed for natural rather than technological
Table 1: Classification of hypothetical global catastrophic risk scenarios by global spread
mechanisms and critical systems affected. Letters represent eight examples of risk scenarios:
asteroid impact (a), volcanic super-eruption (v), pandemic (natural) (p), ecosystem collapse (e),
nuclear war (n), bioengineered pathogen (b), weaponised artificial intelligence (w),
geoengineering termination shock (g). Cell colour represents number of catastrophic scenarios
potentially compromising the critical system globally via the spread mechanism (grey: no likely
disruption, light pink: one scenario, dark pink: two scenarios, red: three or more scenarios).
Critical systems with an identical vulnerability profile to these risk scenarios have been omitted
for brevity, indicated by ellipses (see Figure 1 for the full list of systems).
ACCEPTED MANUSCRIPT

7
disasters, it considers many of the potential human factors that influence
resilience and vulnerability to an unfolding disaster. We take a similar approach
here, and identify potentially fragile areas in the global risk prevention and
mitigation system (Figure 3). Rather than aiming for comprehensiveness or
exclusivity, it highlights that understanding these interdependent and complex
human factors requires input from a wide range of disciplines beyond the
natural sciences.
For instance, short-term thinking and a limited focus constitute cognitive biases
affecting risk perception and management on the individual and institutional
level (as studied in psychology and behavioural economics); unresolved political
conflicts and competing ethical notions of justice undermine international
cooperation and burden-sharing on the institutional and supra-institutional level
(as studied in e.g. law, philosophy and political science).
Some risks (e.g. natural pandemics) are already the focus of well-developed
institutional systems (e.g. the World Health Organization), robust research
activity and technical know-how. For GCRs from emerging technologies,
however, the institutional mix and a research agenda are only just becoming
established. Conventional disaster response (e.g. recovery and compensation),
and even newer, comprehensive strategies (e.g. the “build back better” principle
adopted in some countries post-disaster) are inadequate for addressing threat
scenarios where there is limited reaction time and no second chance. For these
cases, we need a novel framework that is at least as interdisciplinary as the
SFDRR, but moves away from uni-dimensional, natural hazards and instead
addresses complex, anthropogenic risks, which are far more likely to cause a
Figure 3 Levels and dimensions of prevention and mitigation factors moderating global
catastrophic risks.
ACCEPTED MANUSCRIPT

8
severe global catastrophe (Rees, 2003). In particular, we have to focus on the
prevention and mitigation of multidimensional risk scenarios that involve
cascades of socio-technological, natural-technological (“natech”) and
technological-natural disasters.
As we confront emergent technological GCR scenarios, lessons can be learnt from
previous smaller disasters. An instructive recent case of a multi-dimensional
disaster scenario, albeit of local scope, is the Fukushima Dai’ichi nuclear
accident, which laid bare failures at the interface of natural, scientific,
technological, socioeconomic, legal and political realms. One such failure was the
supervision of Japan’s nuclear industry by the very same authorities that were to
promote nuclear technology. Such an institutional setup, aggravated by cognitive
biases (e.g. groupthink) in a sector with revolving doors to the regulator, was
lacking adequate incentive structures, and was destined to result in conflicts of
interest and regulatory capture. The international science and policy community
therefore has the opportunity and the responsibility to co-create better risk
prevention and mitigation systems, by engaging with researchers in the social
sciences and humanities.
In principle it is possible to create a table that would expand on Table 1 to
include the third dimension described here, i.e., prevention and mitigation
failures. Such a table is, however, difficult to produce in practice, as the scenarios
it helps us distinguish between are more fine-grained than those classified in
Table 1. They are subcategories of these scenarios. For example, in Table 1 we
classified “natural pandemic” as a single scenario, yet from a disaster policy and
risk reduction perspective there is a clear difference between a pandemic that
emerged due to underinvestment in veterinary surveillance, and a pandemic that
emerged due to accidental release from a research laboratory. These scenarios
can be further subdivided through the precise failures that allow the pandemic
risk to materialise. If we consider just the accidental release scenario, we would
start from the grid items occupied by ‘p’ in Table 1, which highlight intersections
of the critical systems undermined by pandemic, such as anatomical systems,
and the spread mechanisms for pandemic, which naturally include biological
replicators but are also affected by anthropogenic networks as well as air- and
water-based dispersal. To these we would add a third dimension, that would
highlight all the prevention and mitigation failures potentially involved in
accidental release, from failures of individual skill or risk perception, through
institutional failures including malformed incentives, or insufficient staffing and
resources, to supra-institutional failures of insufficient monitoring and
enforcement.
5 Intended use of the classification system
In this section, we illustrate three key ways the classification system could
potentially be used, although more may be discovered as the system is expanded
and updated.
ACCEPTED MANUSCRIPT

9
The first potential use is to prioritise risk reduction efforts. As can be seen in
Table 1, scenarios with significantly different primary causes could manifest
their GCR potential through a similar mechanism. For example, asteroid impact-,
volcanic super-eruption-, and nuclear war scenarios all feature a risk of
significant reduction of inbound solar radiation, disrupting food security and
potentially leading to mass starvation. Not only does this draw attention to
systems that are vulnerable to multiple hazards, but it also suggests there is
value in considering these scenarios together in research and policy contexts,
rather than thinking about them in isolation. For example, if accounting for
volcanic super-eruptions, asteroid impacts and nuclear wars together, one might
seriously consider risk management strategies that are robust to all scenarios,
such as alternative food production systems to withstand the multi-year “winter”
that might follow (Denkenberger & Pearce, 2015). While this does not preclude
investment in nuclear disarmament or asteroid deflection, it demonstrates that
alternative food policies may warrant more attention than first thought.
In addition to the challenge of securing food under reduced solar radiation, the
classification framework highlights other areas that warrant further attention as
potentially occurring from a range of threats. These include: how to manage the
proliferation of potentially dangerous technologies, how we would function if
human contact was restricted during a pandemic spread, and how we might
make critical digital systems resilient to disruption by error or malice. The value
of the classification system in highlighting potentially compatible risk reduction
strategies is visualised in Table 2.
ACCEPTED MANUSCRIPT

10
Table 2 Classification of risk reduction strategies by global spread mechanisms and critical systems
affected. Letters represent six examples of risk reduction strategies: asteroid deflection (A), digital
resilience (D), food production through non-photosynthetic processes (F), limiting human contact during a
pandemic (L), nuclear disarmament (N), restrictions on the diffusion of risky technologies (R). Cell colour
represents number of risk reduction strategies addressing possible critical system failure and its global
spread via the mechanism (grey: not addressed, light green: one strategy, green: two strategies, dark green:
three or more strategies). Critical systems with an identical benefit profile from these strategies have been
omitted for brevity, indicated by ellipses (see Figure 1 for the full list of systems).
While expansion of this table into the third dimension of prevention and
mitigation failures is beyond the scope of the current paper, we foresee that the
creation of such an expansion, in a dynamic and collaborative fashion as
described below, will have the same benefits as Tables 1 and 2. That is, it could
be used to focus attention on prevention and mitigation failure categories that
affect a range of GCR scenarios (e.g. better risk communication tools). While
policy relevance to multiple risks does not directly entail higher priority for an
intervention (as matters of probability, effectiveness and cost need to be taken
into account), it could indicate the value of a comprehensive cross-risk analysis,
to paint a more complete picture of the value of a proposed intervention.
The second potential use for the classification system lies in creating a live
reference list of expertise for different risk scenarios. Our attempt to carve out
categories in each dimension based on different academic domains should
provide a quick index of the academic disciplines that are essential to “have at
the table” when researching a specific risk scenario. Such an index could prove
useful for policy makers who take responsibility for certain risk domains, or
when an emerging risk is unfolding and an interdisciplinary team needs to be
assembled in a hurry. This potential use underscores the importance of including
the third dimension, which points to relevant academic disaster management
ACCEPTED MANUSCRIPT

11
expertise outside the natural sciences.
The third potential use for the system is as a tool to highlight highly uncertain or
neglected corners of the GCR possibility space, and guide research efforts
towards these corners, in the hope of discovering unknown unknowns. The
combinatorial nature of the classification systems provides a natural way of
progressing from well-known systems and mechanisms to a vast and as-yet
largely unexplored space of possible GCRs. Admittedly, even an exhaustive
exploration of all possible GCR scenario configurations within the current
classification system would not provide a guarantee against “black swans”, but it
can certainly foster a fuller understanding of the threats we face.
6 Where to next?
The classification framework presented above is dynamic, spanning a broad
range of disciplines and reflecting a dense web of interacting variables along
three dimensions: where critical systems are vulnerable to GCRs, how threats
might spread globally, and how attempts to prevent or mitigate these threats
might fail due to human factors. To successfully maintain awareness and
organise the plethora of knowledge around GCRs we need to meet the following
challenges:
1. collect, aggregate and digest information from highly distributed
knowledge networks, overcoming communication barriers and delays;
2. update regularly the classification of GCR scenarios as knowledge
advances, and as technology shapes—or is poised to shape—the relevant
domains.
Meeting these challenges requires a combination of strategies. It would be
sensible to populate a classification framework using a group elicitation
approach, calling on experts in different critical systems, global reach
mechanisms and mitigation approaches to produce short summaries containing
signposts to evidence in their fields that would be relevant to GCRs. Such
summaries would then be aggregated in a central repository. A group of multi-
domain experts could serve as editors to make sure efforts are coordinated,
language is harmonised and appropriate for an interdisciplinary audience, and
credit is attributed appropriately. Similar, successful repositories for other
disciplines already exist and could provide inspiration (Zalta, 2016; Wolfrum,
2017). The evolving classification system, when part of a knowledge synthesis
effort, could offer a visual way to communicate the current state of knowledge
(McKinnon, 2015).
As the frontiers of knowledge and innovation expand, so too does the horizon of
our possible futures. The framework outlined here could both inform, and be
informed by, different ‘foresight’ tools (Cook et al., 2014). It may be a useful tool
for generating scenarios that help us explore and prepare for new risks,
emerging trends and key uncertainties. Scenarios can then be characterised in
more detail and monitored using horizon scanning (Sutherland & Woodroof,
ACCEPTED MANUSCRIPT

12
2009; van Rij, 2010; Amanatidou, 2012), another tool in the ‘foresight’ suite.
Structured horizon scanning methods could be useful to scan for the early
signals of a scenario unfolding, or simply to update the classification framework
with information on new discoveries, innovation, theories and data produced by
the scientific community.
Globalisation and technology are advancing at a rapid pace, and it is difficult to
appraise the ever-changing landscape of risks. In order for research into new,
potentially disruptive technologies to proceed responsibly, and to better
anticipate how interacting threats may unfold across our globe, the state of
knowledge around risks and potential risk mitigation measures needs to be
transparent, organised and updateable. We hope that the classification
framework outlined in this paper will facilitate the communication between
disciplines that such an endeavour needs.
ACCEPTED MANUSCRIPT

13
References
Amanatidou, E. et al. (2012). On concepts and methods in horizon scanning.
Science and Public Policy 39, 208–221.
Asimov, I. (1981). A Choice of Catastrophes: The Disasters That Threaten Our
World. Fawcett Columbine, New York.
Baum, S.D. and Handoh, I.C. (2014). Integrating the planetary boundaries and
global catastrophic risk paradigms. Ecological Economics 107, 13-21.
Baum, S.D. and Tonn, B.E. (eds.) (2015). Confronting future catastrophic threats
to humanity [Special issue]. Futures 72, 1-96.
Bostrom, N. (2002). Existential Risks: Analyzing Human Extinction Scenarios and
Related Hazards. Journal of Evolution and Technology 9.
Bostrom, N. and Ćirković, M.M. (eds.) (2008). Global Catastrophic Risks. Oxford
University Press.
Coburn, A.W., Bowman, G. Ruffle, S.J. Foulser-Piggott, R. Ralph, D. and Tuveson,
M. (2014). A Taxonomy of Threats for Complex Risk Management.
Cambridge Risk Framework series. Centre for Risk Studies, University of
Cambridge.
Cook, C.N., Inayatullah, S., Burgman, M.A., Sutherland, W.J., Wintle, B.A. (2014).
Strategic foresight: how planning for the unpredictable can improve
environmental decision-making. Trends in Ecology & Evolution 29, 531–541.
Cotton-Barratt, O., Farquhar, S., Halstead, J., Schubert, S. and Snyder-Beattie, A.
(2016). Global Catastrophic Risks 2016. Global Challenges Foundation.
Denkenberger, D. and Pearce, J.M. (2015). Feeding Everyone No Matter What.
Elsevier.
IRDR (2014). Peril Classification and Hazard Glossary.
McKinnon, M.C., Cheng, S.H., Garside, R., Masuda, Y.J., and Miller, D. C. (2015).
Sustainability: Map the evidence. Nature 528(7581), 185-187.
Morrison, K.D. (2006). Failure and how to avoid it. Nature 440, 752-754.
Posner, R.A. (2004). Catastrophe: risk and response. Oxford University Press.
Rees, M.J. (2003). Our Final Hour. Basic Books.
Rockström, J. et al. (2009). Planetary boundaries: exploring the safe operating
space for humanity. Ecology and Society 14(2), 32.
Steffen, W. et al. (2015). Planetary boundaries: Guiding human development on a
changing planet. Science 347(6223), 1259855.
Sutherland, W.J. and Woodroof, H.J. (2009). The need for environmental horizon
scanning. Trends Ecol Evol 24, 523-527.
Tonn, B.E. and MacGregor, D. (2009). Human Extinction [Special issue]. Futures
41, 673-774.
Turchin, A. (2015). How many X-Risks for Humanity? This Roadmap has 100
Doomsday Scenarios. Institute for Ethics and Information Technology
https://ieet.org/index.php/IEET2/more/turchin20150623 [last accessed:
19 May 2017]
UNISDR (2015). Sendai Framework for Disaster Risk Reduction 2015–2030.
van Rij, V. (2010). Joint horizon scanning: identifying common strategic choices
and questions for knowledge. Science and Public Policy 37, 7-18.
Wolfrum R. (ed.) (2017). The Max Planck Encyclopedia of Public International
Law. Oxford University Press. http://opil.ouplaw.com/home/EPIL [last
ACCEPTED MANUSCRIPT

14
accessed: 19 May 2017].
Zalta, E.N. (ed.) (2016). Stanford Encyclopedia of Philosophy. Stanford University.
https://plato.stanford.edu/ [last accessed: 19 May 2017].
ACCEPTED MANUSCRIPT

15
ACCEPTED MANUSCRIPT