Decisions, Games and Logic 2015

The Eighth Workshop in Decisions, Games and Logic (DGL) will be hosted by the Department of Philosophy, Logic and Scientific Method at the London School of Economics on 17-19 June 2015. The DGL workshop series aims to bring together graduate students, post-docs and researchers from philosophy, economics and logic working on formal approaches to rational individual and interactive decision making.

Tutorials

Keynotes

Causal vs. Evidential Decision Theory Roundtable

Contributed Papers

This year, we invite submissions from graduate students, post-docs and other early career researchers in the fields of Decision Theory, Game Theory, Logic and Formal Philosophy. Preference will be given to conceptual/foundational work in these fields and to interdisciplinary approaches.

Submissions to DGL2015 can be made in the two following tracks:

Submit your paper Contact Us Download CFP

Important Dates

Day 1 (17 June)

Day 2 (18 June)

Day 3 (19 June)

On June 19 - 20, 2015 the Department of Philosophy, Logic and Scientific Method will host a workshop on Decision Making under Severe Uncertainty. This event will take place under the joint auspices of the Managing Severe Uncertainty project based at the LSE and the DUSUCA project based at GREGHEC (HEC Paris, CNRS). The speakers for this event will be announced soon. All participants to Decisions, Games and Logic 2015 are cordially invited to attend.

Tutorial

Keynote

Causal vs. Evidential Decision Theory Roundtable

Contributed Papers

Aidan Kestigan (Carnegie Mellon)

Reliable Decision Making in Epistemic Democracies

Epistemic democrats argue that democracies are preferable to other political structures at least in part because democratic procedures such as majority voting are the most reliable methods for selecting good outcomes. In order to prove that democratic majority voting is the most reliable political decision procedure, some epistemic democrats have employed a particular formal result: the Condorcet Jury Theorem (hereafter CJT). This paper provides a voting theoretic argument against epistemic democratic theories that use the CJT. I consider the kinds of democratic voting systems that would meet the epistemic democrat's reliability standard. Using modeling tools from game theory and voting theory, I argue that epistemic democrats can only justify the use of a small class of idealized majority voting procedures, and not the more general class of procedures that are usually considered to be “democratic.” This, I argue, is a failure of the theory.

Colin Elliot (Tilburg)

Testing opinions through decisions: betting odds and sincere degrees of belief

In subjective probability, degrees of belief are measured by observing agents' decisions. We propose a new critique of the betting definition of degrees of belief. There, a bookie proposes a bet to an agent over the truth of a verifiable proposition. The betting odds the agent accepts are then taken as her sincere degree of belief about the object of the bet. Crucially, the bookie decides the direction of the bet after hearing the agent's odds. We argue that it is reasonable, then, that the agent should think about how the bookie, her opponent, decides to do this: it makes all the difference between losing or winning money if the event occurs. We show that it is enough for the agent to make very general assumptions about the bookie, in order to observe discrepancies between sincere credences and declared betting odds.

Sara Aronowitz (Michigan)

Why an ideal agent may have stochastic belief generation mechanisms

This paper presents some considerations in favor of employing methods for randomly choosing one's credence from a set distribution, and discusses two objections to such an approach. The first is that stochastically generated beliefs don't count as beliefs, since they are chosen rather than generated as a direct response to evidence. The second is that even if randomly chosen beliefs count as beliefs, there's no evidential situation in which random choice would do as well or better than traditional credence (either imprecise or precise). I argue that sampling one's credences from a distribution captures the advantages of imprecise credence views while improving the agent's ability to learn.

Tutorial

Keynote

Contributed Papers

Paolo Galeazzi (ILLC) and Michael Franke (Tubingen)

Smart Transformations: or, the Evolution of Choice Principles

Evolutionary game theory classically investigates which behavioral patterns are evolutionarily successful in a single game. More recently, a number of contributions have studied the evolution of preferences instead: which subjective conceptualizations of a game's payoffs give rise to evolutionarily successful behavior in a single game. Here, we want to extend this existing approach even further by asking: which general patterns of subjective conceptualizations of payoff functions are evolutionarily successful across a class of games. In other words, we will look at evolutionary competition of payoff transformations in “meta-games”, obtained from averaging over payoffs of single games. Focusing for a start on the class of 2x2 symmetric games, we show that regret minimization can outperform payoff maximization if agents resort to a security strategy in case of radical uncertainty.

Aydin Mohseni (Carnegie Mellon)

The Limits of Equilibrium Concepts in Evolutionary Game Theory

Within the modeling framework of evolutionary game theory static equilibrium concepts adapted from rational choice game theory are employed to identify the probable outcomes of dynamic evolutionary processes. Over the last several decades, results have emerged in the literature demonstrating limitations of each of the proposed equilibrium concepts. We present a comprehensive story circumscribing the shortcomings of the Nash equilibrium (NE), evolutionarily stable strategy (ESS), neutrally stable strategy (NSS), and evolutionarily stable set (ESSet) concepts. We argue that these results rely on an implicit notion of evolutionary significance, and propose a novel account of evolutionary significance. We show how this formulation brings clarity to assessments of the success and failure of equilibrium concepts. We demonstrate that even under the most favorable assumptions regarding the underlying dynamics and conditions for stability—those of replicator dynamics, and asymptotic stability—each equilibrium concept falls short. Each one is simultaneously too weak and too strong. That is, each picks out population states that we would not consider plausible outcomes of an evolutionary process and each fails to pick out states that constitute eminently plausible evolutionary outcomes.

Jurgis Karpus (King's College London) and Mantas Radzvilas (LSE)

Team Reasoning and a Rank-Based Function of Team's Interests

Orthodox game theory is sometimes criticized for its failure to single out intuitively compelling solutions in certain scenarios. The theory of team reasoning provides a resolution in some such cases by suggesting a shift in decision-makers' mode of reasoning from individualistic to reasoning as members of a team. For an explicit function of team's goals a reference is sometimes made to the maximization of the average of individuals' personal payoffs. We criticize this suggestion due to its reliance on interpersonal comparisons of players' payoffs and its potential advocacy of a complete sacrifice of individuals' personal interests for the benefit of other members of a team. In this paper we propose an alternative, rank-based function of team's interests that does not rely on interpersonal comparisons of payoffs and precludes the possibility of a self-sacrifice. We discuss its predictions using a number of examples and suggest a few possibilities for further research.

Tutorial

Keynote

Contributed Papers

Kevin Dorst (MIT)

Epistemic Logic as Epistemology: Paradoxes of Higher-Order Evidence

Higher-order evidence is evidence about what one’s evidence suggests, such as when one is told that one has been slipped a reasoning-distorting drug after doing a logic puzzle. Intuitively, there are systematic level-bridging norms between such levels of evidence, for example: one shouldn’t believe p while believing that this belief is irrational. However, recently some theorists have denied that there are any such systematic connections. It is widely known that doing so saddles them with certain paradoxical results. What has not yet been fully appreciated about the view is that (1) these known paradoxes may be even more problematic than is commonly realized, and (2) they are not the end of the story, for there are at least three further paradoxes of the view. In short, it is far harder to deny the existence of level-bridging norms than many have thought.

Adam Bjorndahl (Carnegie Mellon)

Language-based Games

We introduce language-based games, in which utility is defined over descriptions in a given language. By choosing the right language, we can capture psychological games and reference-dependent preferences. Of special interest are languages that can express only coarse beliefs (e.g., the probability of an event is “high” or “low”, rather than “the probability is .628”): by assuming that a player’s preferences depend only on what is true in a coarse language, we can resolve a number of well-known paradoxes in the literature, including the Allais paradox. Despite the expressive power of this approach, we show that it can describe games in a simple, natural way. Nash equilibrium and rationalizability are generalized to this setting; Nash equilibrium is shown not to exist in general, while the existence of rationalizable strategies is proved under mild conditions on the language. (joint work with Joseph Halpern and Rafael Pass)

Erich Rast (New University of Lisbon)

Making up One's Mind: From Values to Value Judgments

In this talk, I outline a theory of value that highlights the connection between possibly incomplete and conflicting value representations and how these are applied in a decision situation or when particulars are compared. The proposal is based on defeasible rules whose antecedent conditions reflect factual requirements that any value judgment needs to fulfill and whose succedents are comparisons between alternatives formulated on the basis of utilities and the preference relations these represent. It is argued that such defeasible preconditions provide an elegant way of dealing with the context sensitivity of `better than' comparisons with multiple attributes and helps explaining why many value disputes are factual despite being based on seemingly subjective preferences. Many arguments about values do not concern the values themselves, but rather the veridicality of corresponding preconditions and the adequacy of the contextual resolution process within a specific decision situation or in a situation when an agent makes general value comparison between particulars.

The poster session will take place on 18 June from 12-14 in OLD.328. Poster presenters will have 5 minutes to deliver a slides-based presentation of their research. Afterwards lunch will be served and the audience will have the opportunity to discuss with the speakers in an informal setting.

Poster Presenters

Matthew Mandelkern (MIT)

The Logic of Speaker Presupposition and the Semantic Representation of Uncertainty

The language that we use to express our degree of certainty in a given proposition— the language of epistemic modality—has been taken to represent a fundamental challenge for the orthodox view of conversational dynamics according to which an assertion, if accepted, updates the conversation’s context set by intersecting it with the asserted content. In light of this challenge, a substantial number of philosophers and linguists have rejected the traditional view of conversational dynamics. I argue that this is too hasty: we can represent the sentences in question as ordinary propositions which update the context set by intersection. This approach is only viable, however, if we extend the logic of speaker presupposition so that the substitution instances of a certain schema are all valid. The schema in question says, roughly, that if the speaker presupposes that the context set has a certain structural property (a property having to do with its compatibility, entailment, or measure-theoretic properties), then the set of worlds compatible with the speaker’s presuppositions indeed has that structural property. I argue that we should adopt this schema as a theorem of the logic of speaker presupposition. It does not follow from the logic of speaker presupposition given in Stalnaker (2002), but I argue that it nevertheless follows from Stalnaker’s informal characterization of speaker presupposition in that paper. I propose an extension of Stalnaker’s logic that entails the schema in question; if we adopt this logic, which is independently plausible, we can preserve the attractive traditional view of conversational dynamics.

Jean Baccelli (ENS)

On the axiomatic significance of risk attitudes

When making decisions under risk, one can follow several decision rules. Decision theory is concerned with analyzing those rules, especially with axiomatizing them. It is natural to inquire whether risk attitudes, under their technical definitions, play a role in such analysis. First, this paper provides evidence to the contrary. Using Allais’s paradoxes as illustrations, it shows that decision rules are essentially neutral with respect to risk attitudes. Second, it explains why this neutrality should be qualified and how decision rules are fruitfully analyzed in terms of risk attitudes, as illustrated with a more comprehensive discussion of Allais’s paradoxes. It shows that most decision rules can be characterized by the type of conditional risk attitude variation they allow, and distinguished by the extent to which they impose the generalization of risk attitudes. By assessing the axiomatic significance of risk attitudes, this paper contributes to the epistemology of decision theory.

Josh Turkewitz (Florida State)

A Problem for Probabilism

James M. Joyce’s “A Nonpragmatic Vindication of Probabilism” presents a powerful argument for the claim that it is epistemically rational to conform one’s degrees of partial belief to the laws of probability theory. Importantly, his argument is not based on the prudential costs of probabilistic inconsistency (such as the unavoidable vulnerability to clever bookies aware of this inconsistency). Instead he argues that (just as with full beliefs) partial beliefs can be more or less accurate based on how well they ‘fit the facts’ and gives a comprehensive and compelling account of what it is for a partial belief to ‘fit the facts’ better than another. I argue that an important measure of accuracy (what has been called verisimilitude) is not captured by Joyce’s system, and movement along this standard might raise or lower the accuracy of a system of partial beliefs without regard to the axioms of probability. Crucially, movement along this standard could result in a probabilistically inconsistent system of partial beliefs being more accurate than a probabilistically consistent system of partial beliefs, so Joyce’s vindication is not wholly successful.

Nicolas Wüthrich and Wulf Gaertner (LSE)

Evaluating Competing Theories Via a Common Language of Qualitative Verdicts

Kuhn (1977) claimed that several algorithms can be defended to select the best theory based on epistemic values such as simplicity and accuracy. In a recent paper, Okasha (2011) argued that no theory choice algorithm exists which satisfies a set of intuitively compelling conditions. In this paper, we propose a solution to avoid the impossibility result. Based on Gaertner and Xu (2012), we suggest reconstructing theory choice with the help of a general scoring rule defined over a set of qualitative intervals for every epistemic value. This aggregation method yields a complete and transitive ranking and the rule satisfies the Arrovian conditions within a cardinal setting. We show that our procedure can be extended to capture the aggregation across different scientists.

Simon Hewitt (Birkbeck)

Plural Logic and Logicism

This short talk develops a version of logicism about arithmetic using a plural logic, motivated by the position that numbers are plural properties. Equipped with an abstraction principle for ordered pairs, we define equinumerosity, and go on to derive an implementation of the Dedekind-Peano axioms. We offer philosophical reflection on the result, drawing attention to the possibility of nonstandard plural models and hinting towards the construction of number systems beyond the naturals.

Boris Babic (Michigan)

Frequency Encoding Credences: An Imprecise Guide to Direct Inference

The problem of direct inference takes the following form: (i) we have a large general population, (ii) from which a sample is taken, and (iii) we are asked to formulate a credence about some individual in the general population on the basis of the relative frequencies observed in the sample. I call degrees of belief that result from this process frequency encoding credences. I argue that in most cases of philosophical interest the dominant approach (strict calibration) enjoins us to adopt credences that either fail to reflect available information or encode unavailable information. I then offer an alternative rule for eliciting frequency encoding credences (distortion adjusted calibration) that better reflects an agent's total evidence at any given time

Conference Venues

LAK.206

On the second floor of the Lakatos Building (LAK).

LAK.G01C

On the ground floor of the Lakatos Building (LAK).

OLD.328

On the third floor of the Old Building (OLD).

The Venue (Party on 18 June)

The basement of Saw Swee Hock Student Centre (SAW).

Ciao Bella (Dinner on 17 June)

86-90 Lamb's Conduit St, Bloomsbury, WC1N 3LZ.

Loch Fyne (Dinner on 18 June)

2-4 Catherine Street, London, The City, WC2B 5YJ.

Travel Information

Located in central London, LSE is easily accessible by a range of public transport including tube, rail and bus. Cycling and driving to LSE are also options.

Tube/Underground stations

• Holborn (Piccadilly and Central lines): Approximately five minutes away.
• Temple (District and Circle lines): Approximately five minutes away.
• Charing Cross (Jubilee, Northern and Bakerloo lines): Approximately ten minutes away.

British Rail stations

• Waterloo: Approximately 10-15 minutes away.
• Charing Cross: Approximately 10 minutes away.
• Blackfriars: Approximately 15 minutes away.
• St Pancras International: Take the Piccadilly Line from King's Cross to Holborn. Serves Eurostar.

Buses

Buses that stop on or near the Aldwych are: 1, 4, 6, 9, 11, 13, 15, 23, 26, 59, 68, X68, 76, 77a, 91, 139, 168, 171, 172, 176, 188, 243, 341 and 521.

Each bus stop should show which buses stop there and their frequency. On the front of the bus the final destination will be given. It may also show the names of the main stops on its route.

Car

LSE is located within London's congestion charging zone. See London Congestion Charging Homepage for details of how to pay it. There are only a few parking meters around the LSE campus, mainly near Lincoln's Inn Fields. The closest NCP parking is on Parker St off Drury Lane.

Cycling

There are bicycle racks on campus at St Clement's Building; in Grange Court; opposite the main entrance on Houghton Street; and in the courtyard behind St Philips Building (entrance on Sheffield Street). There are eight high-security cycle racks in the basement area between St Philips north and south buildings. Anyone using these racks must use a padlock to ensure that their bike is secure, not a D lock or chain. Please note that these secure racks are down a flight of steel steps.

The LSE campus now also hosts two Barclays Cycle Hire docking stations on Houghton Street and Portugal Street. For more information about these please see the Barclays Cycle Hire (Transport for London website) web page. For further information on cycling in London please see the Cycles page on the Transport for London website. Please note that bicycle theft does take place in central London, including LSE campus.

General travel information

London is divided into travel zones which will dictate the cost of your journey by underground (tube), rail or bus. The best way to keep travel costs down is to buy a standard adult Oyster card from any tube station or location where you see the Oyster card logo. These allow you discounted journeys by tube, Docklands Light Railway (DLR), bus and some National Rail services within greater London. For more details please see the Transport for London website. This is especially useful for information about public transport in London, with details about all bus and tube journeys.

Accommodation

DGL does not have discounted rates with any of the hotels around the campus. However, the following two are often used by speakers visiting the LSE: Club Quarters, Goodenough College. Alternatively, you can look for more affordable accommodation on Booking.com and Airbnb

Childcare

The LSE Nursery offers childcare to any visitor to the LSE. However, they have limited places and they cannot guarantee that they will be able to accommodate anyone at anytime. In consequence if you wish to make use of this facility please contact us as soon as possible in order to inquire about their availability.

The organisers wish to express their gratitude for the support they have received from the following institutions: British Society for the Philosophy of Science, The Centre for Philosophy of Natural and Social Science, The Department of Philosophy, Logic and Scientific Method and the LSE Choice Group.

To get in contact with us, please send us an email at: Philosophy.Probability@lse.ac.uk