Speaker: Golbon Zakeri
Affiliation: UoA Engineering Science
Title: Electricity market modelling, economics and analytics
Date: Wednesday, 20 Aug 2014
Time: 4:00 pm
Location: CAG17/114-G17 (Commerce A)

Over the past 2 decades there has been a major shift to meet the electricity needs of various countries and jurisdictions through markets. We will start by describing issues common to the vast majority of electricity systems and reasons that rationalised the move to electricity markets in developed countries. We will then discuss issues that arise from a transition to an electricity market with a particular focus on the NZ electricity market. This is a rich source of mathematical modelling, economics and analytics problems. We will lay out some of the more interesting problems that we have tackled and go in more depth to explore consequences of the introduction of renewables and our proposed solutions.

This talk is targeted towards members with varied backgrounds.

Everyone welcome!

Speaker: Jeremy Seligman
Affiliation: The University of Auckland (Philosophy)
Title: Secret tweets and network discovery
Date: Wednesday, 6 Aug 2014
Time: 5:00 pm
Location: Room 405, Engineering (403)

You are a secret agent with a secret S that you would like to transmit to a fellow agent a unobtrusively using a very public network like Twitter. Any information you tweet will be received by your followers on the network. You correctly assume that they will send the message on to their followers (retweet it) if and only if it does not conflict with any information they already possess. With luck, your message will be tweeted through the network until it eventually reaches a. Under what conditions is it possible for you to convey S to a in this way, without other agents in the network learning this information? Clearly,you cannot tweet S itself, but if, for example, a is the only agent to know that K then the message `if K then S’ may work, if there is a suitable path from you to a. To know whether you can succeed or not and what to tweet, you need to know something about the network and the information already possessed by the other agents. But you can learn something about this with a test tweet. If, for example, you know that you have two followers b and c and only b believes P and then you tweet the message `not P’ then if, after a certain length of time, someone tweets P to you, you know that there is a loop back to you via c. This talk will report on recent joint work on these and similar questions with Mostafa Raziebrahimsaraei.

Speaker: Matthew Ryan
Affiliation: Department of Economics, UoA
Title: Binary Stochastic Choice under Risk or Uncertainty
Date: Tuesday, 3 Jun 2014
Time: 5:00 pm
Location: Room 412, Science Centre (303)

Economists usually model choice as deterministic, via preference relations, though occasionally — and usually for econometric convenience — choice is allowed to be stochastic. Psychologists, on the other hand, typically model choice behaviour as intrinsically stochastic. In psychophysics, for example, it is common to model the probability of choosing one option over another (in a binary choice problem) as an increasing function of the difference in “utility” stimuli associated with the options. This is called a “strong utility representation” (SUP) for the binary choice probabilities.

These models of binary stochastic choice generate numerous interesting mathematical problems. This talk will introduce a small sample. The main focus will be on binary choice problems in which each option is a “lottery”, with risky or uncertain value. Given a specification of choice probabilities for all possible pairs of lotteries, under what conditions (on these probabilities) does there exist a SUP? What if we additionally require that the utility scale exhibit particular properties, such as linearity?

Speaker: Patrick Girard, Valery Pavlov and Mark Wilson
Affiliation: The University of Auckland
Title: Experimental study of influence in social networks
Date: Tuesday, 20 May 2014
Time: 5:00 pm
Location: OGGB level 0. DECIDE, The UoA Business Decision-Making Lab (formerly Lab 04

We present an overview of work in progress using the DECIDE lab (http://hfbeltran.wix.com/decide), part of a larger project on threshold models of influence. Our pilot study has yielded several interesting hypotheses for future investigation. We welcome audience suggestions for follow-up work.


Speaker:     Kerry Manson
Affiliation: The University of Auckland
Title:       Power Indices and their Real-World Application
Date:        Tuesday, 8 Apr 2014
Time:        5:00 pm
Location:    Room 412, Science Centre (303)

It has been recognised that standard, a priori power indices do not describe political power
well in many real-world situations. This lowers our ability to usefully incorporate the effects of
power distribution into political analyses. Methods of a posteriori power measurement which
take into account restrictions on the set of allowable coalitions are examined and developed. A
Hasse diagram representation of games is used to give a new method for adapting any a priori
Banzhaf index to a restricted coalition structure. A modified Banzhaf index is further analysed, and the
conditions for manipulation of this index by players are given.

The talk is based on Kerry’s Honours dissertation written in 2013.

Speaker:     Sergey Ozernikov
Affiliation: The University of Auckland
Title:       Public-Key Infrastructure: trust metrics in the Web of Trust
Date:        Tuesday, 25 Mar 2014
Time:        5:00 pm
Location:    Room 412, Science Centre (303)

Public Key Insrastructure (PKI) is an arrangement that provides its users with means for confident and effective utilisation of public-key cryptography. An overview of non-hierarchical example of PKI – Pretty Good Privacy (PGP) – will be given, which uses the concept of a Web of Trust – a structure where any user can act as a certificate authority and assert validity of other users’ certificates.

A user A of PGP before sending a message to another user X must calculate the validity of the other user’s certificate using the public information about the network and their private information about the trustworthiness of the users on the certification paths from A to X. This is normally done on the basis of the so-called trust metric. Trust metrics and decision rules based on them are a field of active research – it is unlikely that anyone will invent a single perfect decision rule since there are many conflicting desiderata. In particular, it is desirable that such a decision rule
– be immune to various attacks;
– be easily computable;
– satisfy nice normative properties (axioms).

This is an introductory talk on this subject. No specific knowledge of cryptography will be assumed. Social networks people are especially invited.

Speaker:     Adam Clearwater
Affiliation: The University of Auckland
Title:       The single-crossing property on a tree
Date:        Tuesday, 11 Mar 2014
Time:        5:00 pm
Location:    Room 412, Science Centre (303)

We generalize the classical single-crossing property to single-crossing property on trees and obtain new ways to construct the so-called Condorcet domains which are sets of linear orders which possess the property that every profile composed from those orders have transitive majority relation. We prove that for any tree there exist profiles that are single-crossing on that tree; moreover, that tree is minimal in this respect for at least one such profile. Finally, we provide a polynomial-time algorithm to recognize whether or not a given profile is single-crossing with respect to some tree. We also show that finding winners for Chamberlin-Courant rule is polynomial  for profiles that are single-crossing on trees.

This paper is a product of Adam’s Summer Scholarship project. The research was conducted jointly with Clemens Puppe (KIT, Germany) and Arkadii Slinko.

Speaker:     Arkadii Slinko
Affiliation: The University of Ackland
Title:       Swensson’s theorem and its failed generalisations
Date:        Tuesday, 25 Feb 2014
Time:        5:00 pm
Location:    Room 412, Science Centre (303)

Swensson’s theorem is one of the many impossibility theorems in mathematical economics. These impossibility theorems give you an idea of what is possible and what is impossible to achieve. Imagine that you have to allocate state houses to families who need them and that those families have preferences on the set of houses. What should the allocation mechanism be? Swensson (1999) proved that if we impose just three simple desirable properties on the allocation mechanism we will be left only with a serial disctatorship under which a random queu (permutation) will be chosen and each family will be asked to choose their house when their term in the queue comes.

During Piotr Skowron’s visit in January we tried to generalise Swensson’s theorem to more general class of mechanisms called social assignment rules. We produced a bunch of counterexamples instead. In my talk I will prove Swensson’s theorem and present our counterexamples.

Speaker: Andy Philpott (Department of Engineering Science)
Topic: A Primer on Supply-Function Equilibrium
When: 2:30-3:30, Tuesday 29 October
Where: Room 5115, OGGB
Abstract: Supply function equilibrium models arise when agents offer a schedule of prices and quantities to an auction for a single divisible good. They were first developed in the setting of treasury auctions, but have become useful models for studying auctions of electricity, where uncertainty plays a key role. This talk will attempt to give an elementary account of supply function equilibrium, focusing on the mathematics underlying the model.

Slides are available.

Speaker: Matthew Ryan (Department of Economics)
Topic: Belief Functions (Part II)
When: 2:30-3:30, Tuesday 22 October
Where: Room 5115, OGGB

Belief functions generalise the notion of probability by relaxing additivity, while retaining a weaker property called infinite monotonicity. Belief functions allow us to quantify beliefs in a manner which is sensitive to the strength of the evidential support. I’ll focus on how to update such beliefs; more generally, how to perform statistical inference when the prior is described by a belief function. Many puzzles and problems arise when considering the issue of updating/inference. This talk will be informal (i.e., ill-prepared!) and will raise questions rather than provide answers.