João Balsa

Tax compliance in a simulated heterogeneous multi-agent society

We consider an individualised approach to agent behaviour in an application to the classical economic problem of tax compliance. Most economic theories consider homogeneous representative agent utilitarian approaches to explain the decision of complying or not with tax payment. However, a heterogeneous and individualised account of decision can be considered to explain certain apparently irrational behaviours. Ideas such as trust and peer perception may have a key influence in individual decisions, and thus transform the global results for society. In this paper, we apply the agent view of rationality to economic decisions and define a territory to be explored by agent technology and social simulations. We conclude that the multi-agent view can provide powerful results which might lead to significant economic policy implications.

Tactical exploration of tax compliance decisions in multi-agent based simulation

Tax compliance is a field that crosses over several research areas, from economics to machine learning, from sociology to artificial intelligence and multi-agent systems. The core of the problem is that the standing general theories cannot even explain why people comply as much as they do, much less make predictions or support prescriptions for the public entities. The compliance decision is a challenge posed to rational choice theory, and one that defies the current choice mechanisms in multi-agent systems. The key idea of this project is that by considering rationally-heterogeneous agents immersed in a highly social environment we can get hold of a better grasp of what is really involved in the individual decisions. Moreover, we aim at understanding how those decisions determine tendencies for the behaviour of the whole society, and how in turn those tendencies influence individual behaviour. This paper presents the results of some exploratory simulations carried out to uncover regularities, correlations and trends in the models that represent first and then expand the standard theories on the field. We conclude that forces like social imitation and local neighbourhood enforcement and reputation are far more important than individual perception of expected utility maximising, in what respects compliance decisions.

e*plore v.0: Principia for Strategic Exploration of Social Simulation Experiments Design Space

Years ago, we addressed the issue of methodological procedures to develop the design of cognitive agents tuned to real problems, inserting them into a context where experimentation could have a meaningful outcome in terms of the original problems posed. Since then we have been building mechanisms and frameworks for mind design in multiagent systems. We proceeded with the evaluation of such systems through simulation that was many times exploratory. In many of those experiments, the evaluation of the deep meaning of outcomes was inherently complex, challenging the researchers and even the research questions.

Force Versus Majority: A Comparison in Convention Emergence Efficiency

In open societies such as multi-agent systems, it is important that coordination among the several actors is achieved efficiently. One economical way of capturing that aspiration is consensus: social conventions and lexicons are good examples of coordinating systems, where uniformity promotes shared expectations of behavior and shared meanings. We are particularly interested in consensus that is achieved without any central control or ruling, through decentralized mechanisms that prove to be effective, efficient, and robust. The nature of interactions and also the nature of society configurations may promote or inhibit consensual emergence. Traditionally, preference to adopt the most seen choices (the majority option) has dominated the emergence convention research in multi-agents, being analyzed along different social topologies. Recently, we have introduced a different type of interaction, based on force, where force is not defined a priori but evolves dynamically. We compare the Majority class of choice update against Force based interactions, along three dimensions: types of encounters, rules of interaction and network topologies. Our experiments show that interactions based on Force are significantly more efficient (fewer encounters) for group decision making.

Context Switching versus Context Permeability in Multiple Social Networks

In social life, actors engage simultaneously in several relations with each other. The complex network of social links in which agents are engaged is fundamental for any realistic simulation of social life. Moreover, not only the existence of multiple-modality paths between agents in a simulation, but also the knowledge that those agents have about the quality and specificity of those links are relevant for the decisions the agents make and the consequences they have both at an individual and at a collective level. Each actor has a context in each of the relations that are represented as support of a simulation. We build on previous work about permeability between those contexts to study the novel notion of context switching. By switching contexts, individuals carry with them their whole set of personal characteristics to a different relation, while abandoning the previous one. When returning to the original context, all previous links are resumed. We apply these notions to a simple game: the consensus game, in which agents try to collectively achieve an arbitrary consensus through simple locally informed peer-to-peer interactions. We compare the results for context switching with results from simulating the simple game and the game with context permeability.

Agents that collude to evade taxes

We explore the link between micro-level motivations leading to and being influenced by macro-level outcomes to study the complex issue of tax evasion. If it is obvious why there is a benefit for people who evade taxes, it is less obvious why people would pay any taxes at all, given the the small probability of being caught, and the small penalties involved. We use exploratory simulation and progressively deepening models of agents and of simulations to study the reasons behind tax evasion. We have unveiled some relatively simple social mechanisms that can explain the compliance numbers observed in real economies. We claim that simulation with multiple agents provides a strong methodological tool with which to support the design of public policies.

Tax compliance through MABS: the case of indirect taxes

Multi-agent systems can be used in social simulation to get a deeper understanding of complex phenomena, in such a way that predictions can be provided and tested, and policies can be designed in a solid individually grounded manner. Our aim is explore this link between micro-level motivations leading to and being influenced by macro-level outcomes in an economic setting where to study the complex issue of tax evasion. While it is obvious why there is a benefit for people who evade taxes, it is less obvious why people would pay any taxes at all, given the small probability of being caught, and the small penalties involved. Our research program uses exploratory simulation and progressively deepening models of agents and of simulations to study the reasons behind tax evasion. We have unveiled some relatively simple social mechanisms that can explain the compliance numbers observed in real economies. We claim that simulation with multiple agents provides a strong methodological tool with which to support the design of public policies.

A Distributed Approach for a Robust and Evolving NLP System

We present in this paper some aspects concerning the design and implementation of an architecture that is the basis for the development of a natural language processing system that, besides the obvious goal of building some computational representation (at a desired level) of the input, has two main objectives: to be robust and to evolve. To be robust in the sense that the non recognition of some input should not block the system but, instead, should lead the system to an automatic recovery process. To evolve, so that when some incompleteness/incorrectness is detected (or suspected) during a recovery process, the component responsible for the mistake should be updated accordingly, so that in future analogous situations the system can perform better. In order to achieve this goal we propose the definition of a distributed architecture.

Overcoming incomplete information in NLP systems — Verb subcategorization

A new methodology for overcoming incomplete information available for current natural language parsers will be presented in this paper. Although our aim is more ambitious, in this paper, we will focus on incomplete descriptions of the subcategorization classes of verbs and will sketch a proposal for overcoming the same problem for other syntactic categories. We assume a hierarchical multi-agent system architecture where each bottom-layer agent has a specialised knowledge (perspective) about the problems a given feature (e.g. verb subcategorization) of a syntactic category may have. Each agent has a declarative description of those problems and can find better solutions for the parsing problem once it has got an explanation for it. We are assuming logic based diagnosis agents. Each theoretically plausible hypothesis found must then be statistically validated. The pruning obtained and the ordering of validated hypothesis leads then to a learning problem that must be solved in order to enable a natural evolution of parsers (and their lexicons).

Integrating Client Profiling in an Anti-money Laundering Multi-agent Based System

Continuing previous work by the authors, where an Anti-Money Laundering (AML) agent-based system was introduced, we now provide some detail on one of the elements of this system—the learning component. The system we are developing focuses on how a financial institution, a bank, can obtain better results in AML initiatives. More specifically, we’re trying to improve the suspicious transaction signaling process and the subsequent final decision. For this, it is critical to model client behavior, having a clear definition of the different client profiles. Having available a real world data set of bank transactions, we explain in this contribution how some data-mining techniques were used in order to build the needed client profiles, and how the results obtained can be integrated in the system.