Lent Term 2015
Thursday, 25 February, 2:00
Gian-Kasper Plattner (Swiss Federal Research Institute, and former Head of IPCC WGI Technical Support Unit)
IPCC 2013/2014: Assessing the Science of Climate Change
Dr Plattner's talk will be followed by two short research presentations by MSU project members on the topic of IPCC uncertainty assessment. Approximate schedule:
2:00 -- 3:30 Gian-Kasper Plattner slides (pdf)
3:45 -- 4:15 Roman Frigg and Nicolas Wuethrich
4:15 -- 4:45 Richard Bradley and Casey Helgeson
Thursday, 10 March, 2:00
Roberto Buizza (European Centre for Medium-Range Weather Forecasts)
Signals and noise, hawk-moths and butterflies: Weather prediction in a world of uncertainties
ABSTRACT. Ensemble-based, probabilistic systems provide (at least up to now) the most effective way to predict the weather taking into account all relevant sources of uncertainty, to extract predictable signals from sometimes noisy single forecasts. They help us living and dealing with both the butterfly effect (sensitive dependence to initial condition errors; see Lorenz 1963 J. Atm. Sci.) and the hawk-moth effect (sensitive dependence to model approximations; see e.g. Frigg et al 2014 Phy. Of Sci.), and issue skilful predictions despite the chaotic nature of the atmosphere.
In the first part of my talk I will discuss lessons learned in weather prediction, where in the past 25 years we have witnessed a paradigm shift from a single-forecast to a probabilistic approach, whereby a range of possible weather scenarii are predicted using an ensemble. To ensure the reliability of these probabilistic predictions, we have been designing these ensembles to simulate all relevant sources of forecast uncertainties, linked to uncertainties in the initial conditions (the knowledge of the current state of the system) and the model approximations.
I will then discuss how we are applying a similar approach to generate an ensemble of coupled ocean-land-atmosphere reanalyses (i.e. 3- dimensional states of the Earth system) of the past century, to have a less-uncertain estimate of the past climate. Finally, I will conclude the seminar by suggesting some possible links between butterflies, hawk- moths and reliable signals, and economics and forward guidance.
Michaelmas Term 2015
Thursday, 1 October, 2:00
Discussion: 'How probabilities reflect evidence'
This meeting will be a reading-group style discussion of the article: HOW PROBABILITIES REFLECT EVIDENCE, James Joyce (2005). Philosophical Perspectives, 19. paper
Thursday, 22 October, 2:00
Casey Helgeson (LSE) and Katie Steele (LSE)
Integrating Theory and Practice in Deep Uncertainty Decision Science
ABSTRACT. How do applied tools for deep uncertainty decision support (e.g., Robust Decision Making, Decision Scaling, Info-Gap, Adaptive Pathways) relate to the more exact and mathematised decision models of philosophy and economics? And how can theoreticians from the latter tradition best contribute to the development of applied decision support? We discuss possible answers and illustrate with a close look at an approach to simplifying state spaces that is often used in robust decision making.
Thursday, 3 November, 2:00
Hauke Riesch (Brunel University London)
Levels of Uncertainty
ABSTRACT. There exist a variety of different understandings, definitions, and classifications of risk, which can make the resulting landscape of academic literature on the topic seem somewhat disjointed and often confusing. In this chapter, I will introduce a map on how to think about risks, and in particular uncertainty, which is arranged along the different questions of what the different academic disciplines find interesting about risk. This aims to give a more integrated idea of where the different literatures intersect and thus provide some order in our understanding of what risk is and what is interesting about it. One particular dimension will be presented in more detail, answering the question of what exactly we are uncertain about and distinguishing between five different levels of uncertainty. I will argue, through some concrete examples, that concentrating on the objects of uncertainty can give us an appreciation on how different perspectives on a given risk scenario are formed and will use the more general map to show how this perspective intersects with other classifications and analyses of risk.
Thursday, 10 December, 2:00
Alessandro Tavoni (LSE Grantham Institute)
Climate change games in the lab: economic experiments on bargaining and mitigation
Related blog posts by Alessandro:
Summer Term 2015
Thursday, 30 April, 2:00
Bernard Sinclair-Desgagné (HEC Montreal) and Pauline Barrieu (LSE CATS)
Economic Policy when Models Disagree
ABSTRACT. This paper introduces a general approach to conceive public policy when there is no consensual model of the situation of interest. This approach builds on one basic attribute of rational policymakers -- namely their ability to appraise their experts’ scenarios and forecasts -- and uses only one normative criterion: that the value to policymakers of a remedy’s projected outcomes meets their willingness to escape the current situation. Unlike the methods currently put forward in the literature, it does not need -- but is compatible with -- a representative policymaker’s objective function (as in the ambiguity aversion literature), a reference model (as in robust control theory) or some prior probability distribution over the set of supplied scenarios (as in Bayesian model-averaging). It can handle situations where policy decisions are made through collective deliberations and voting. Policies constructed in this manner are shown to be effective, robust, simple and precautionary in a precise and intuitive sense.
Working paper | Slides [pptx]
Thursday, 7 May, 2:00
Magda Osman (Queen Mary University)
What processes enable us to control uncertainty?
ABSTRACT. The presentation will start with a round trip of psychological research in the domain of dynamic control. For the most part the literature has focused heavily, and unnecessarily, on weak empirical demonstrations of unconscious processes that are thought to underpin control behaviours that are recruited to manage uncertain situations. The presentation will then go on to discuss the latest advances in the field. This work suggest that the way we are able to exert control in situations in which information is highly impoverished is when we: (1) can assert a sense of agency over our circumstances, and (2) when we identify signals in the environment that we judge to be reliable indicators of past performance (e.g., reward information), that can, in turn, be used to determine desirable future outcomes. Slides [pptx]
Thursday, 14 May, 2:00
Erica Thompson (LSE CATS), Roman Frigg (LSE), and Casey Helgeson (LSE) will facilitate a discussion of their working paper: Expert Judgment for Climate Change Adaptation (email C.Helgeson@lse.ac.uk for the paper).
ABSTRACT. Climate change adaptation is largely a local matter, and adaptation planning can benefit from local climate change projections. Such projections are typically generated by accepting climate model outputs in a relatively uncritical way. We argue, based on the IPCC’s treatment of model outputs from the CMIP5 ensemble, that this approach is unwarranted and that subjective expert judgment should play a central role in the provision of local climate change projections intended to support decision-making.
Thursday, 4 June, 2:00
Work-in-progress presentations by (mostly) PhD students (LSE Philosophy) - PART 1: Nicolas Wuethrich, Casey Helgeson
Conceptualizing uncertainty: An assessment of the latest IPCC’s uncertainty framework
ABSTRACT. We are facing severe uncertainties regarding the phenomenon of climate change. To address these uncertainties, the Intergovernmental Panel on Climate Change has introduced a new version of its framework for communicating uncertainty which involves a confidence and a likelihood metric to qualify findings. In this paper, I critically assess this framework. First, I look at the meta-documents which explain the uncertainty framework and argue that there are substantial conceptual issues which need attention. Secondly, I explore how the uncertainty framework is put into practice and show that the conceptual problems of the framework manifest themselves in concrete practical problems for the authors and readers of the assessment report. Based on these observations, I suggest, thirdly, improvements for the framework. In particular, I argue that the confidence metric needs to be constructed in a different way which involves the clarification of the two sub-metrics agreement and evidence.
Low confidence in extreme probabilities
ABSTRACT. Qualitative notions of 'confidence' are used by the IPCC and other organisations to indicate the strength of evidence behind a claim. Sometimes the claims themselves are probabilistic, meaning that the overall characterisation of uncertainty is spread across the claim and the confidence qualification. This divided uncertainty characterisation raises questions like: Do all combinations of probability and confidence make sense? Kandlikar et al. (2005), and Risbey and Kandlikar (2007) say, without much argument, that low confidence in extreme probabilities (near zero or one) are incoherent. I present some considerations for and against their position.
Thursday, 11 June, 2:00
Matthew Adler (Duke University)
Climate Change and Prioritarianism
ABSTRACT. Prioritarianism is the ethical view that gives greater weight to well-being changes affecting individuals at lower well-being levels. This view is influential both in moral philosophy, and in theoretical work on social choice—where it is captured by a social welfare function summing a concave transformation of individual utilities. However, prioritarianism has been largely ignored by scholarship on climate change. This talk compares utilitarianism and prioritarianism as frameworks for evaluating climate policy. It reviews the distinctive normative choices that are required for the prioritarian approach: for example, the specification of a ratio scale for utility, the degree of concavity of the transformation function (i.e., the degree of social inequality aversion), and the choice between “ex ante” and “ex post” prioritarianism under conditions of risk. It also examines some of prioritarian’s salient implications, for example the social cost of carbon, the social discount rate, and the “dismal theorem.” Slides [pptx]
Friday and Saturday, June 19-20.
Workshop on Decision Making under Severe Uncertainty
Thursday, 2 July, 2:00
Work-in-progress presentations by (mostly) PhD students (LSE Philosophy) - PART 2: Silvia Milano, Thomas Rowe
Lent Term 2015
Thursday, 15 January, 2:00
Simon Dietz (LSE Grantham Institute), presenting work co-authored with Nicoleta Anca Matei (Joint Research Centre, European Commission, and University of Verona)
Spaces for agreement: A theory of Time-Stochastic Dominance and an application to climate change
ABSTRACT. Many investments involve both a long time-horizon and risky returns. Making investment decisions thus requires assumptions about time and risk preferences. Such assumptions are frequently contested, particularly in the public sector, and there is no immediate prospect of universal agreement. Motivated by these observations, we develop a theory and method of finding ‘spaces for agreement’. These are combinations of classes of discount and utility function, for which one investment dominates another (or ‘almost’ does so), so that all decision-makers whose preferences can be represented by such combinations would agree on the option to be chosen. The theory is built on combining the insights of stochastic dominance on the one hand, and time dominance on the other, thus offering a non-parametric approach to inter-temporal, risky choice. We go on to apply the theory to the controversy over climate policy evaluation and show with the help of a popular simulation model that, in fact, even tough carbon emissions targets would be chosen by almost everyone, barring those with arguably ‘extreme’ preferences.
Thursday, 22 January, 2:00
Hykel Hosni (LSE)
Symmetric vs asymmetric extensions of classical expected utility - PART 1
ABSTRACT (for parts 1 & 2). J Maynard Keynes is often recognised as one of the early skeptics about the adequacy of probability as an all-encompassing measure of uncertainty. In a much quoted passage Keynes characterises his understanding of "uncertain knowledge" as follows:
"The sense in which I am using the term is that in which the prospect of a European war is uncertain [...]. About these matters there is no scientific basis on which to form any calculable probability whatever. We simply do not know. Nevertheless, the necessity for action and for decision compels us as practical men to do our best to overlook this awkward fact and to behave exactly as we should if we had behind us a good Benthamite calculation of a series of prospective advantages and disadvantages, each multiplied by its appropriate probability, waiting to be summed."
I will illustrate informally the key ideas underlying some prominent models of uncertainty quantification and decision-making which can be seen as addressing Keynes's challenge of providing us with a "good Benthamite calculation" in the face of severe uncertainty. In doing this I will suggest that it is useful to group and compare models according to whether they assume that "knowledge" and "ignorance" can be treated symmetrically.
Thursday, 26 February, 2:00
Richard Bradley (LSE) and Casey Helgeson (LSE), presenting work co-authored with Brian Hill (HEC, Paris)
Decision and Climate Change Assessments
ABSTRACT. The Intergovernmental Panel on Climate Change (IPCC) has adopted official guidelines for communicating uncertainties attached to the scientific findings that appear in their assessment reports. One novel feature of their approach is the use of ‘confidence’, a second-order qualification that indicates the level of scientific understanding behind a claim. So far, it is unclear how findings communicated in this format can be used to inform rational decisions. We explore how the IPCC guidelines fit together with a new, confidence-based normative model of decision making (Hill, 2013, ‘Confidence and Decision’, Games and Economic Behavior).
Thursday, 12 March, 2:00
Hykel Hosni (LSE)
Symmetric vs asymmetric extensions of classical expected utility - PART 2
See 22 Jan. for ABSTRACT. Recommended Reading | Slides [pdf]
Thursday, 19 March, 2:00
Elizabeth Baldwin (LSE Grantham Institute)
Climate Policy with Incomplete Knowledge
ABSTRACT. The economic effects of climate change, and the costs of any policy response, are very uncertain. Models ought to reflect this. The `minimal' theory for decision-making under Knightian uncertainty is is a model of incomplete beliefs and tastes. I define a `dismal' situation in which different assumptions, all plausible, give rise to large estimated costs which differ dramatically. I claim this is the correct way to think about climate change. I then discuss the implications for climate policy, in particular re-assessing the question of `prices versus quantities' Weitzman (1974) in the context of incomplete preferences. This analysis provides new `wrong target' effects, which supplement the original result. The `dismal theorem' of Weitzman (2009) provides an example of my `dismal' situation, but as my broader framing shows, it may have led to too narrow a focus on `very unlikely, very catastrophic' events.
Michaelmas Term 2014
Thursday, 30 October, 2:00
Luc Bovens (LSE) and Bryan Roberts (LSE)
Maximising the Chance of Doing a Best Action: Jill and John Revisited
ABSTRACT. We compare two decision rules viz. (ME) Maximise Expectations and (MCBA) Maximise the Chance of doing a Best Action. ME enjoins us to choose the action which maximises expected utility. MCBA enjoins us to choose the action that maximises the ex post chance of having done a best action. Jackson (Ethics, 1991) argued that MCBA is an unacceptable rule for moral decision-making. His Jill-and-John counter example involves a choice set containing more than two elements and with dependencies between the outcomes of the actions. We examine how both rules behave in a simple binary choice with independent actions. We explore how these decision rules may be more and less reasonable depending on one’s attitudes towards regret minimisation and on the scale of utility measurement. Nonetheless, what remains disconcerting about the MCBA is that it violates separability and that MCBA choosers may become more risk averse when there is less at stake.
Thursday, 27 November, 2:00
Brian Hill (HEC, Paris)
Confidence as a Source of Deferral
ABSTRACT. One apparent reason for deferring a decision – abstaining from choosing, leaving the decision open to be taken by someone else, one’s later self, or nature – is for lack of sufficient confidence in the relevant beliefs. This paper develops an axiomatic theory of decision in situations where a costly deferral option is available that captures this source of deferral. Drawing on it, a preliminary behavioural comparison with other accounts of deferral, such as those based on information asymmetry, is undertaken, and a simple multi-factor model of deferral – involving both confidence and information considerations – is formulated. The model suggests that incorporating confidence can account for cases of deferral that traditional accounts have trouble explaining.
Thursday, 4 December, 2:00
Richard Bradley (LSE)
Pooling Expert Opinion
We will look at the question of how to combine expert opinion in situations in which they disagree in their probabilistic assessments of one or more variables. The main aim will be to survey the existing literature in order to establish what are the most promising lines of research. Everyone should look at the Genest and Zidek survey (focus on sections 3, 4 and 6), which we will use as the basis for discussion. This paper by Dietrich and List is also useful.
Thursday, 11 December, 2:00
Casey Helgeson (LSE)
Climate Change Uncertainties and Expert Elicitation
This talk will introduce the topic of expert-elicited knowledge as an input to risk management and policy decision making, and survey applications in the field of climate/global change. Topics to include: elicitation methodology, whether to aggregate expert opinions, expert elicitation versus `expert judgment' in IPCC assessments, and the role of expert elicitation in climate services. The main aim is to identify areas for research. Slides (Prezi) | Extended Bibliography (pdf)
Summer Term 2014
Thursday, 1 May, 11:00
Discussion of IPCC Working Group I Summary for Policymakers
The Intergovernmental Panel on Climate Change (IPCC) recently released the newest report from Working Group I, the group tasked with assessing 'the physical scientific aspects of the climate system and climate change.' At this meeting we will discuss the Summary for Policymakers (SPM) document that is based on the full report (1500 pages) of Working Group I. WGI SPM
Thursday, 8 May, 2:00
Hykel Hosni (LSE and Scuola Normale Superiore)
Derivatives and systemic risk: A qualitative analysis
Thursday, 22 May, 2:00
Discussion of IPCC WGII and WGIII Summaries for Policymakers
Working Group II of the Intergovernmental Panel on Climate Change (IPCC) addresses impacts, adaptation, and vulnerability; Working Group III addresses mitigation. We will discuss both Summaries for Policymakers at this meeting, with emphasis on WGIII. A quick overview of WGII will be presented, followed by an open-ended discussion of WGIII. Please read the WGIII SPM before the meeting.
Thursday, 5 June, 10:30 - 3:40
Managing Severe Uncertainty project day
The purpose of this event is to discuss and shape the future research directions of the MSU project. Project members will present sketches of their research plans, and guests from the Grantham Research Institute on Climate Change and the Environment, and from the Centre for the Analysis of Timeseries will talk about their own research and/or make suggestions about promising directions of research.
MORNING SESSION, 10:30 -- 12:30. Simon Dietz (Grantham), Charlotte Werndl, Tom Rowe, Dave Stainforth (Grantham and CATS), Katie Steele, Casey Helgeson, Richard Bradley.
LUNCH BREAK, 12:30 -- 1:30.
AFTERNOON SESSION, 1:30 -- 3:10. Erica Thompson (CATS), Alex Voorhoeve, Hykel Hosni, Antony Millner (Grantham), Roman Frigg, Silvia Milano.
Thursday, 26 June, 2:00
A double header of shorter, work-in-progress research presentations, from Alex Voorhoeve (LSE) and Roman Frigg (LSE).
Alex Voorhoeve (co-authors: Ken Binmore, Lisa Stewart and Arnaldur Stefansson)
Framing and Ambiguity Aversion
ABSTRACT: I shall report on a new series of experiments on behaviour in the face of ambiguity. Contrary to other research, we find little effect of framing on people's choices under ambiguity. In particular: 1. We find no significant evidence of a switch from ambiguity-averse behaviour for gains to ambiguity-loving behaviour for losses; 2. We find only modest effects of making ambiguity especially salient. Indeed, the model that best fits our data is one in which subjects primarily choose in accordance with the Sure-Thing Principle, with a modest tendency to depart from this principle in an ambiguity-averse direction.
Evidence-Based Policy in the Face of Uncertainty?
ABSTRACT: I review the prospects of evidence based policy in the face of severe uncertainty. I criticise approaches that produce probabilities based on model outputs and suggest that actionable projections should be produced by aggregating suitably elicited expert judgment.
Lent Term 2014
Thursday, 23 January, 2:30 - 4:30
Decision Theory Masterclass (1 of 3), Richard Bradley (LSE)
This masterclass will provide an introduction to the main theory of decision making under uncertainty, covering the Savage and Anscombe-Aumann framework's, the Ellsberg paradox and the debate over the quantification of uncertainty, theories of decision under ignorance and perhaps some of the early alternatives to Savage. The class will not assume any prior knowledge of decision theory but will cover the basics at very high speed. Slides [pdf]
Thursday, 30 January, 2:30 - 4:30
Ken Binmore (Bristol and UCL, Economics and Philosophy)
Rational Decisions in Large Worlds
ABSTRACT: Savage denied that Bayesian decision theory applies in large worlds. A minimal extension of Bayesian decision theory to a large-world context assigns upper and lower probabilities to ambiguous or uncertain events—those to which a single probability cannot be assigned. The orthodox Hurwicz criterion evaluates an ambiguous or uncertain event as a weighted arithmetic mean of its upper and lower probability. The ambiguity or uncertainty aversion reported in experiments on the Ellsberg paradox is then explained by assigning a larger weight to the lower probability of winning than to the upper probability. This paper reviews a possible framework for assessing both the arithmetic Hurwicz criterion and a geometric variant. It then argues that only the case of equal weights satisﬁes appropriate consistency requirements—a result that accords with our own experiments on the Ellsberg paradox. However, the geometric Hurwicz criterion allows another and more severe interpretation of uncertainty aversion that persists even in the case of equal weights.
Thursday, 13 February, 2:30 - 4:30
Decision Theory Masterclass (2 of 3), Katie Steele (LSE)
This 2nd Decision Theory Masterclass pursues a topic that is related but somewhat independent of Decision Theory Masterclasses 1 and 3. The first part will introduce the standard decision theory (expected utility) model for determining the ‘value of information’, or in other words, how much we should be willing to pay (whether in time, money, brainpower or other resources) to pursue new information/evidence, given the decision problems that we face. Philosophical applications of this model will be discussed, and its limitations explored. The second part will shift to the context of severe uncertainty, and what role considerations of the ‘value of information’ may play in debates about rational decision-making in this context.
Thursday, 27 February, 2:30 - 4:30
Charlotte Werndl (LSE)
On Defining Climate and Climate Change
This talk analyses the main candidates for a definition of climate and climate change. Five desiderata on a definition of climate are presented: it should be empirically applicable, it should correctly classify different climates, it should not depend on our knowledge, is should be applicable to the past, present and future and it should be mathematically well-defined. Then five definitions are discussed: climate as distribution over time for constant external conditions, climate as distribution over time when the external conditions vary as in reality, climate as distribution over time relative to regimes of varying external conditions, climate as the ensemble distribution for constant external conditions, and climate as the ensemble distribution when the external conditions vary as in reality. The third definition is novel and is introduced as a response to problems with existing definitions. The conclusion is that most definitions encounter serious problems and that the third definition is most promising. Paper [pdf]
Thursday, 6 March, 2:30 - 4:30
Decision Theory Masterclass (3 of 3), Richard Bradley (LSE)
In this third masterclass I will be presenting the main 'post-Bayesian' models of decision making under uncertainty (ambiguity), including Maximin EU, the 'smooth' model of ambiguity and Bewley's theory of choice under incompleteness. In order to understand the background for these models I will look briefly at the older literature on decision making under ignorance and present the Anscombe-Aumann framework within which much of the contemporary literature works. Slides [pdf]
Michaelmas Term 2013
Thursday, 24 October, 2:30 - 4:30
Climate Models Masterclass (1 of 3), Roman Frigg (LSE)
The three masterclasses are in effect a crash course in the natural science aspects of climate change. Everybody wishing to understand the basic physics behind climate change and the use of climate models is welcome to attend; no prior knowledge is presupposed. The first lecture explains the earth's energy balance and introduces basic concepts such as the greenhouse effect, radiative forcing, time lags, feedback loops, climate sensitivity and climate variability. The second lecture reviews the empirical evidence for climate change and discusses some sceptical claims. The third lecture introduces climate models and discusses what kinds of uncertainties attach to them. Slides [pdf]
Thursday, 14 November, 2:30 - 4:30
Climate Models Masterclass (2 of 3), Roman Frigg (LSE)
See Oct 24 meeting for description. Slides [pdf]
Thursday, 21 November, 3:00 - 4:30
The Classification of Uncertainty, Richard Bradley (LSE)
This meeting will be devoted to a discussion of how different kinds of uncertainty should be classified. Note the time change: 3:00 - 4:30.
Thursday, 5 December, 2:30 - 4:30
Climate Models Masterclass (3 of 3), Roman Frigg (LSE)
See Oct 24 meeting for description. Slides [pdf]
Thursday, 12 December, 2:30 - 4:30
Discussion of forthcoming papers by Nicholas Stern