Thesis research at the University of Minnesota (1974-1975) on Games, Goals, and Bounded Rationality.
-
Leigh Tesfatsion, "Two Essays on Individual Choice"
(pdf,2.5MB),
PhD Thesis, Department of Economics, 76-15,007 University of Minnesota, Mpls., December 1975.
-
The key idea of this thesis is that boundedly-rational decision-makers can supplement partial contingency plans with goals,
thus permitting actions to be undertaken in desired directions despite the absence of complete information. An expected utility model permitting goal-directed action choice is formulated and axiomatized for concrete demonstration.
Work at the University of Minnesota (1974-5) and University of Southern California (1975-1984) on
Criterion Filtering.
-
The following question arises for sequential decision-making under uncertainty: Can the associative mapping between decisions and expected payoffs be directly updated, without recourse to the updating of probabilities?
-
Criterion Filtering (CF) provides an affirmative answer to this question. CF is a learning algorithm that permits the direct updating of expected utility on the basis of sequentially realized utility outcomes, conditional on prior (initial) utility assessments. The CF algorithm thus provides a "Bayes' Rule for Utility".
-
-
Criterion Filtering (CF): A Dual Approach
to Bayesian Inference and Adaptive Control
Work at University of Southern California (1975-1990) with applied mathematician Bob Kalaba on Flexible Least Squares.
- Flexible Least Squares (FLS) is a multicriteria optimization method for model specification. The goal of the FLS method is to identify the "Pareto frontier" of all efficiently estimated models conditional on: (i) a given theory; (ii) a given data set; and (iii) a designated collection of goodness-of-fit metrics. FLS does not require the ad-hoc specification of conventional stochastic assumptions for "residual error terms" that in fact arise from deterministic model misspecification.
-
Flexible Least Squares (FLS): A Multicriteria Optimization Method for Model Specification
Work at USC (1975-1990) with applied mathematician Bob Kalaba on Adaptive Computation.
-
An adaptive computation method adapts to the problem at hand rather than requiring the problem to be adapted to the method.
Bob Kalaba and I developed adaptive computation methods for the solution of various types of nonlinear problems.
-
For example, we developed an adaptive homotopy solution algorithm able, in runtime, to detect and avoid regions where calculations are ill-conditioned due to nearby singularities or bifurcation points. This adaptive capability is achieved by replacing the standard homotopy continuation parameter, moving in a pre-set manner from 0 to 1 along the real line, with a "smart agent" able to construct and traverse an adaptively determined path from 0+0i to 1+0i in the complex plane. The smart agent decides the direction and length of its next step, given its current state, by solving a multi-criteria optimization problem requiring a trade-off between two criteria: (i) maintain a short path length from 0+0i to 1+0i; and (ii) avoid regions in the complex plane where ill-conditioning of calculations is detected.
-
Adaptive Computation Methods for Nonlinear Systems;
Work at USC (1975-1990) and ISU (1990-present) on
Modeling of Economies as Open-Ended Dynamic Systems.
- This work includes, for example, the modeling of macroeconomies as DSGL systems, where
DSGL = DSG(~E) + Learning Agents.
-
More precisely, DSGL systems are Dynamic,
Stochastic, and General market-based macroeconomic systems for which decision-making agents have Learning capabilities. Coordination arises endogenously (if at all) in DSGL systems as a result of successive agent interactions, not from the external imposition of equilibrium assumptions.
-
Optimality and Efficiency of
Open-Ended Dynamic Economies
Work at ISU (1990-2001) on the Blending of Game Theory with Matching Theory.
-
Trade network games are explored for concrete demonstration. The sequential choice and refusal of trade partners (matching theory) is modeled, together with the sequential determination of trade strategies (evolutionary game theory). At the start-time for each successive trading period, each trader directs trade offers to preferred trade partners and preferentially accepts or refuses trade offers received from other traders,
based on his current state (data,attributes,methods). Matched traders then participate in a trade process modeled as a two-player game, where each trade partner uses a trade strategy selected to maximize his expected net gain conditional on his current state. At the completion of these trade processes, each trader adaptively updates his current state based on trade outcomes he has experienced or observed during this latest trading period.
-
The Trade Network Game (TNG) Laboratory is open-source demonstration software permitting the beautiful run-time visualization of endogenous trade-network formation among strategic buyers, sellers, and dealers in a sequential trade network game.
-
Endogenous Trade Network Formation
and the TNG Laboratory;
Work at ISU (1990-present) on
Completely Agent-Based Modeling (c-ABM) and
Agent-based Computational Economics (ACE).
-
Scientists and engineers seek to understand how real-world systems work, and how real-world systems could work better. Any modeling approach devised for such purposes must simplify reality. Ideally, however, the modeling approach should be flexible as well as logically rigorous; it should permit model simplifications to be appropriately tailored for the specific purpose at hand.
-
Modeling flexibility and logical rigor have been the two key goals motivating my development of completely Agent-Based Modeling (c-ABM)
and Agent-based Computational Economics (ACE). The
c-ABM modeling approach is a variant of agent-based modeling characterized by seven modeling principles. Any model adhering to these seven modeling principles is a computational laboratory permitting explorations of computational systems in a manner analogous to biological experimentation with cultures in Petri dishes. The ACE modeling approach is a specialization of c-ABM to economic systems.
-
Completely Agent-Based Modeling (c-ABM)
-
Agent-Based Computational
Economics (ACE): Homepage
Work with electric power engineers at ISU (2006-present) on market-based designs for Integrated Transmission and Distribution Systems.
-
The growing reliance of centrally-managed wholesale electric power markets on renewable power resources poses new challenges for these markets. For example, intermittent resources such as wind power and solar power that are not back-stopped by storage can make it more difficult to balance power withdrawal (usage & losses) with power injection (supply) across a transmission grid, a prerequisite for reliable grid operations.
-
These challenges have led to efforts by the U.S. Federal Energy Regulatory Commission -- such as FERC Order 2222 (Final Rule) released in September 2020 -- to increase the participation of distributed power resources in these markets in dispatchable aggregated form. Transactive Energy System (TES) design is a relatively new approach to electric power management that could provide important support for FERC Order 2222 objectives.
-
In response to these trends, our ISU project on
Integrated Transmission and Distribution (ITD) systems has developed
various types of TES designs to facilitate the efficient reliable provision of power and ancillary services from a wide array of ITD power resources in return for appropriate compensation. This ITD TES design research has been supported by the development of agent-based computational platforms representing the salient operations of actual electric power systems.
-
Integrated Transmission and Distribution (ITD)
Project: ISU Homepage
Work at ISU (2011-present) on a Linked Swing-Contract Market Design for Grid-Supported Wholesale Electric Power Markets
-
Good market design begins with an appropriate conceptualization of the products being transacted.
For centrally-managed wholesale electric power markets operating over high-voltage transmission grids, these products take two forms.
- (1) Physically-Covered Insurance: The guaranteed availability of power-generating capabilities for possible central dispatch during future operating periods to protect against "volumetric grid risk," i.e., the possible disruption of grid operations due to real-time imbalance between the injection of power into the grid and the withdrawal and/or inadvertent loss of power from the grid.
-
(2) Real-Time Power-Delivery Performance: The actual delivery of power in response to central dispatch instructions received during successive operating periods to satisfy actual just-in-time customer power demands and grid reliability requirements.
-
In the Wiley/IEEE Press book cited below, a linked swing-contract market design is proposed for grid-supported centrally-managed wholesale power markets to facilitate increased reliance on renewable power together with more active demand-side participation. The proposed swing contracts are two-part pricing contracts, in firm or option form, that permit dispatchable power resources to offer physically-covered insurance and real-time power delivery for successive operating periods in return for appropriate compensation.
-
Dispatchable power resources submit swing contracts into a
collection of linked forward markets M(OP) managed by a system operator for successive future operating periods OP. The system operator determines an optimal contract-clearing solution for each market M(OP) to maximize the expected net benefit to be attained during OP by the M(OP) participants subject to system constraints for OP that include grid reliability constraints.
-
Leigh Tesfatsion (2021), A New Swing-Contract Design for Wholesale Power Markets, 20 Chapters, 288pp., John Wiley & Sons, Inc. (IEEE Press Series on Power Engineering), Hoboken, New Jersey, USA.
(BookReview,pdf),
(FERCTalk,SlideSet,pdf),
(Wiley/IEEEPressBookFlyer,pdf).
Work at ISU (2020-present) on Economics of Grid-Supported Electric Power Markets: A Fundamental Reconsideration.
-
The key conceptual finding of this work to date is that the use of Locational Marginal Pricing (LMP) to price grid-delivered energy in U.S. centrally-managed day-ahead and real-time wholesale power markets operating over high-voltage AC transmission grids cannot be conceptually supported.
The difficulty is a fundamental unit-of-measurement issue that permeates many current studies of grid-supported electric power markets.
-
Leigh Tesfatsion (2023), "Economics of Grid-Supported Electric Power Markets: A Fundamental Reconsideration"
(WPLatestVersion,pdf,1.1MB),
(SlideSet,pdf,1MB),
Economics Working Paper No. 22005, ISU Digital Repository, Iowa State University, Ames, Iowa, Latest Revision: 18 January 2023.
-
Leigh Tesfatsion (2023), "E-Filed Comments and Supporting Document"
(EFiledComments,pdf,1.4MB),
U.S. Federal Energy Regulatory Commission (FERC), Docket AD21-10-000, Modernizing Wholesale Electric Power Market Design," Filed: January 14, 2023.