Skip to main content

Über dieses Buch

This volume focuses on uncovering the fundamental forces underlying dynamic decision making among multiple interacting, imperfect and selfish decision makers.

The chapters are written by leading experts from different disciplines, all considering the many sources of imperfection in decision making, and always with an eye to decreasing the myriad discrepancies between theory and real world human decision making.

Topics addressed include uncertainty, deliberation cost and the complexity arising from the inherent large computational scale of decision making in these systems.

In particular, analyses and experiments are presented which concern:

• task allocation to maximize “the wisdom of the crowd”;

• design of a society of “edutainment” robots who account for one anothers’ emotional states;

• recognizing and counteracting seemingly non-rational human decision making;

• coping with extreme scale when learning causality in networks;

• efficiently incorporating expert knowledge in personalized medicine;

• the effects of personality on risky decision making.

The volume is a valuable source for researchers, graduate students and practitioners in machine learning, stochastic control, robotics, and economics, among other fields.



Chapter 1. Bayesian Methods for Intelligent Task Assignment in Crowdsourcing Systems

In many decision-making scenarios, it is necessary to aggregate information from a number of different agents, be they people, sensors or computer systems. Each agent may have complementary analysis skills or access to different information, and their reliability may vary greatly. An example is using crowdsourcing to employ multiple human workers to perform analytical tasks. This chapter presents an information-theoretic approach to selecting informative decision-making agents, assigning them to specific tasks and combining their responses using a Bayesian method. For settings in which the agents are paid to undertake tasks, we introduce an automated algorithm for selecting a cohort of agents (workers) to complete informative tasks, hiring new members of the cohort and identifying those members whose services are no longer needed. We demonstrate empirically how our intelligent task assignment approach improves the accuracy of combined decisions while requiring fewer responses from the crowd.
Edwin Simpson, Stephen Roberts

Chapter 2. Designing Societies of Robots

We provide a framework to model competition and cooperation within a group of agents. Competition is dealt with through adversarial risk analysis, which provides a disagreement point and, implicitly, through minimum distance to such point. Cooperation is dealt with through a concept of maximal separation from the disagreement point. Mixtures of both problems are used to refer to in-between behavior. We illustrate the ideas with several experiments in relation with groups of robots.
Pablo G. Esteban, David Ríos Insua

Chapter 3. On the Origins of Imperfection and Apparent Non-rationality

Decision making (DM) is a preferences-driven choice among available actions. Under uncertainty, Savage’s axiomatisation singles out Bayesian DM as the adequate normative framework. It constructs strategies generating the optimal actions, while assuming that the decision maker rationally tries to meet her preferences. Descriptive DM theories have observed numerous deviations of the real DM from normative recommendations. The explanation of decision-makers’ imperfection or non-rationality, possibly followed by rectification, is the focal point of contemporary DM research. This chapter falls into this stream and claims that the neglecting a part of the behaviour of the closed DM loop is the major cause of these deviations. It inspects DM subtasks in which this claim matters and where its consideration may practically help. It deals with: (i) the preference elicitation; (ii) the “non-rationality” caused by the difference of preferences declared and preferences followed; (iii) the choice of proximity measures in knowledge and preferences fusion; (iv) ways to a systematic design of approximate DM; and (v) the control of the deliberation effort spent on a DM task via sequential DM. The extent of the above list indicates that the discussion offers more open questions than answers, however, their consideration is the key element of this chapter. Their presentation is an important chapter’s ingredient.
Miroslav Kárný, Tatiana V. Guy

Chapter 4. Lasso Granger Causal Models: Some Strategies and Their Efficiency for Gene Expression Regulatory Networks

The detection of causality in gene regulatory networks from experimental data, such as gene expression measurements, is a challenging problem. Granger causality, based on a vector autoregressive model, is one of the most popular methods for uncovering the temporal dependencies between time series, and so it can be used for estimating the causal relationships between the genes in the network. The application of multivariate Granger causality to the networks with a big number of variables (genes) requires a variable selection procedure. For fighting with lack of informative data, the so called regularization procedures are applied. Lasso method is a well known example of such a procedure and the multivariate Granger causality method with the Lasso is called Graphical Lasso Granger method. It is widely accepted that the Graphical Lasso Granger method with an inappropriate parameter setting tends to select too many causal relationships, which leads to spurious results. In our previous work, we proposed a thresholding strategy for Graphical Lasso Granger method, called two-level-thresholding and demonstrated how the variable over-selection of the Graphical Lasso Granger method can be overcome. Thus, an appropriate thresholding, i.e. an appropriate choice of the thresholding parameter, is crucial for the accuracy of the Graphical Lasso Granger method. In this paper, we compare the performance of the Graphical Lasso Granger method with an appropriate thresholding to two other Lasso Granger methods (the regular Lasso Granger method and Copula Granger method) as well as to the method combining ordinary differential equations with dynamic Bayesian Networks. The comparison of the methods is done on the gene expression data of the human cancer cell line for a regulatory network of nineteen selected genes. We test the causal detection ability of these methods with respect to the selected benchmark network and compare the performance of the mentioned methods on various statistical measures. The discussed methods apply a dynamic decision making. They are scalable and can be easily extended to networks with a higher number of genes. In our tests, the best method with respect to the precision and computational cost turns out to be the Graphical Lasso Granger method with two-level-thresholding. Although the discussed algorithms were motivated by problems coming from genetics, they can be also applied to other real-world problems dealing with interactions in a multi-agent system.
Kateřina Hlaváčková-Schindler, Sergiy Pereverzyev

Chapter 5. Cooperative Feature Selection in Personalized Medicine

The chapter discusses a research support system to identify diagnostic result patterns that characterise pertinent patient groups for personalized medicine. Example disease is breast cancer. The approach integrates established clinical findings with systems biology analyses. In this respect it is related to personalized medicine as well as translational research. Technically the system is a computer based support environment that links machine learning algorithms for classification with an interface for the medical domain expert. The involvement of the clinician has two reasons. On the one hand the intention is to impart an in-depth understanding of potentially relevant ‘omics’ findings from systems biology (e.g. genomics, transcriptomics, proteomics, and metabolomics) for actual patients in the context of clinical diagnoses. On the other hand the medical expert is indispensable for the process to rationally constrict the pertinent features towards a manageable selection of diagnostic findings. Without the suitable incorporation of domain expert knowledge machine based selections are often polluted by noise or irrelevant but massive variations. Selecting a subset of features is necessary in order to tackle the problem that for statistical reasons the amount of features has to be in an appropriate relationship to the number of cases that are available in a study (curse of dimensionality). The cooperative selection process is iterative. Interim results of analyses based on automatic temporary feature selections have to be graspable and criticisable by the medical expert. In order to support the understanding of machine learning results a prototype based approach is followed. The case type related documentation is in accordance with the way the human expert is cognitively structuring experienced cases. As the features for patient description are heterogeneous in their type and nature, the machine learning based feature selection has to handle different kinds of pertinent dissimilarities for the features and integrate them into a holistic representation.
Dietlind Zühlke, Gernoth Grunst, Kerstin Röser

Chapter 6. Imperfect Decision Making and Risk Taking Are Affected by Personality

Classic game theory predicts that individuals should behave as rational agents in order to maximize their gain. In real life situations it is observed that human decision making does not follow this theory. Specific patterns of activity in several brain circuits identified in recent years have been associated with irrational and imperfect decision making. Brain activity modulated by dopamine and serotonin is assumed to be among the main drivers of the expression of personality traits and patients affected by Attention deficit hyperactivity disorder (ADHD) are characterized by altered activity in those neuromodulating circuits. We investigated the effect of fairness and personality traits on neuronal and psychological mechanisms of decision making and risk taking in two sets of experiments based on the Ultimatum Game (UG) and the Investment Game (IG). In the UG we found that Fairness and Conscientiousness were associated with responder’s gain and with event-related potentials (ERP) components Feedback-Related Negativity (FRN) and Late Positive component (LPP). In the IG the sum gained during the risky gambling task were presented immediately after half of the trials (condition “high frequency feedback”, HFFB), while the other half were presented at the end of each block (condition “low frequency feedback”, LFFB). Conscientiousness, Agreeableness and Sincerity influenced latencies of the negative deflection occurring at around 200 ms (N200) and the positive wave peaking at around 250 ms (P250) components. The contingent negative variation (CNV) component was affected in a different way in controls and participants with ADHD as a function of the feedback frequency (HFFB  versus  LFFB). These results clearly show that imperfect decision making and risk taking are affected by personality traits and cannot be accounted by models based on rational computations.
Sarah K. Mesrobian, Michel Bader, Lorenz Götte, Alessandro E. P. Villa, Alessandra Lintas
Weitere Informationen

Premium Partner