Complexity Simplicity Simplexity

“In the midst of order, there is chaos; but in the midst of chaos, there is order”, John Gribbin wrote in his book Deep Simplicity (p.76). In this dialectical spirit, we discuss the generative tension between complexity and simplicity in the theory and practice of management and organization. Complexity theory suggests that the relationship between complex environments and complex organizations advanced by the well-known Ashby’s law, may be reconsidered: only simple organization provides enough space for individual agency to match environmental turbulence in the form of complex organizational responses. We suggest that complex organizing may be paradoxically facilitated by a simple infrastructure, and that the theory of organizations may be viewed as resulting from the interplay between simplicity and complexity.

Business model analysis of Mobitto: How to solve the chicken and egg problem of a multi-sided platform?

A Work Project, presented as part of the requirements for the Award of a Masters Degree in Management from the NOVA – School of Business and Economics

A system’s approach to cache hierarchy-aware decomposition of data-parallel computations

Dissertação para obtenção do Grau de Mestre em Engenharia Informática

A clustering approach for vehicle routing problems with hard time windows

Dissertação para obtenção do Grau de Mestre em Logica Computicional

A semi-automatic approach to code smells detection

Eradication of code smells is often pointed out as a way to improve readability, extensibility and design in existing software. However, code smell detection remains time consuming and error-prone, partly due to the inherent subjectivity of the detection processes presently available. In view of mitigating the subjectivity problem, this dissertation presents a tool that automates a technique for the detection and assessment of code smells in Java source code, developed as an Eclipse plugin. The technique is based upon a Binary Logistic Regression model that uses complexity metrics as independent variables and is calibrated by expert‟s knowledge. An overview of the technique is provided, the tool is described and validated by an example case study.

Planning and verification of multipath routing protocols

Conventionally the problem of the best path in a network refers to the shortest path problem. However, for the vast majority of networks present nowadays this solution has some limitations which directly affect their proper functioning, as well as an inefficient use of their potentialities. Problems at the level of large networks where graphs of high complexity are commonly present as well as the appearing of new services and their respective requirements, are intrinsically related to the inability of this solution. In order to overcome the needs present in these networks, a new approach to the problem of the best path must be explored. One solution that has aroused more interest in the scientific community considers the use of multiple paths between two network nodes, where they can all now be considered as the best path between those nodes. Therefore, the routing will be discontinued only by minimizing one metric, where only one path between nodes is chosen, and shall be made by the selection of one of many paths, thereby allowing the use of a greater diversity of the present paths (obviously, if the network consents). The establishment of multi-path routing in a given network has several advantages for its operation. Its use may well improve the distribution of network traffic, improve recovery time to failure, or it can still offer a greater control of the network by its administrator. These factors still have greater relevance when networks have large dimensions, as well as when their constitution is of high complexity, such as the Internet, where multiple networks managed by different entities are interconnected. A large part of the growing need to use multipath protocols is associated to the routing made based on policies. Therefore, paths with different characteristics can be considered with equal level of preference, and thus be part of the solution for the best way problem. To perform multi-path routing using protocols based only on the destination address has some limitations but it is possible. Concepts of graph theory of algebraic structures can be used to describe how the routes are calculated and classified, enabling to model the routing problem. This thesis studies and analyzes multi-path routing protocols from the known literature and derives a new algebraic condition which allows the correct operation of these protocols without any network restriction. It also develops a range of software tools that allows the planning and the respective verification/validation of new protocols models according to the study made.

Managing innovation complexity: About the co-existence of innovation types

The purpose of this study is to contribute to the changing innovation management literature by providing an overview of different innovation types and organizational complexity factors. Aiming at a better understanding of effective innovation management, innovation and complexity are related to the formulation of an innovation strategy and interaction between different innovation types is further explored. The chosen approach in this study is to review the existing literature on different innovation types and organizational complexity factors in order to design a survey which allows for statistical measurement of their interactions and relationships to innovation strategy formulation. The findings demonstrate interaction between individual innovation types. Additionally, organizational complexity factors and different innovation types are significantly related to innovation strategy formulation. In particular, more closed innovation and incremental innovation positively influence the likelihood of innovation strategy formulation. Organizational complexity factors have an overall negative influence on innovation strategy formulation. In order to define best practices for innovation management and to guide managerial decision making, organizations need to be aware of the co-existence of different innovation types and formulate an innovation strategy to more closely align their innovation objectives.

Essays on bounded rationality: Individual decision and strategic interaction

Economics is a social science which, therefore, focuses on people and on the decisions they make, be it in an individual context, or in group situations. It studies human choices, in face of needs to be fulfilled, and a limited amount of resources to fulfill them. For a long time, there was a convergence between the normative and positive views of human behavior, in that the ideal and predicted decisions of agents in economic models were entangled in one single concept. That is, it was assumed that the best that could be done in each situation was exactly the choice that would prevail. Or, at least, that the facts that economics needed to explain could be understood in the light of models in which individual agents act as if they are able to make ideal decisions. However, in the last decades, the complexity of the environment in which economic decisions are made and the limits on the ability of agents to deal with it have been recognized, and incorporated into models of decision making in what came to be known as the bounded rationality paradigm. This was triggered by the incapacity of the unboundedly rationality paradigm to explain observed phenomena and behavior. This thesis contributes to the literature in three different ways. Chapter 1 is a survey on bounded rationality, which gathers and organizes the contributions to the field since Simon (1955) first recognized the necessity to account for the limits on human rationality. The focus of the survey is on theoretical work rather than the experimental literature which presents evidence of actual behavior that differs from what classic rationality predicts. The general framework is as follows. Given a set of exogenous variables, the economic agent needs to choose an element from the choice set that is avail- able to him, in order to optimize the expected value of an objective function (assuming his preferences are representable by such a function). If this problem is too complex for the agent to deal with, one or more of its elements is simplified. Each bounded rationality theory is categorized according to the most relevant element it simplifes. Chapter 2 proposes a novel theory of bounded rationality. Much in the same fashion as Conlisk (1980) and Gabaix (2014), we assume that thinking is costly in the sense that agents have to pay a cost for performing mental operations. In our model, if they choose not to think, such cost is avoided, but they are left with a single alternative, labeled the default choice. We exemplify the idea with a very simple model of consumer choice and identify the concept of isofin curves, i.e., sets of default choices which generate the same utility net of thinking cost. Then, we apply the idea to a linear symmetric Cournot duopoly, in which the default choice can be interpreted as the most natural quantity to be produced in the market. We find that, as the thinking cost increases, the number of firms thinking in equilibrium decreases. More interestingly, for intermediate levels of thinking cost, an equilibrium in which one of the firms chooses the default quantity and the other best responds to it exists, generating asymmetric choices in a symmetric model. Our model is able to explain well-known regularities identified in the Cournot experimental literature, such as the adoption of different strategies by players (Huck et al. , 1999), the inter temporal rigidity of choices (Bosch-Dom enech & Vriend, 2003) and the dispersion of quantities in the context of di cult decision making (Bosch-Dom enech & Vriend, 2003). Chapter 3 applies a model of bounded rationality in a game-theoretic set- ting to the well-known turnout paradox in large elections, pivotal probabilities vanish very quickly and no one should vote, in sharp contrast with the ob- served high levels of turnout. Inspired by the concept of rhizomatic thinking, introduced by Bravo-Furtado & Côrte-Real (2009a), we assume that each per- son is self-delusional in the sense that, when making a decision, she believes that a fraction of the people who support the same party decides alike, even if no communication is established between them. This kind of belief simplifies the decision of the agent, as it reduces the number of players he believes to be playing against { it is thus a bounded rationality approach. Studying a two-party first-past-the-post election with a continuum of self-delusional agents, we show that the turnout rate is positive in all the possible equilibria, and that it can be as high as 100%. The game displays multiple equilibria, at least one of which entails a victory of the bigger party. The smaller one may also win, provided its relative size is not too small; more self-delusional voters in the minority party decreases this threshold size. Our model is able to explain some empirical facts, such as the possibility that a close election leads to low turnout (Geys, 2006), a lower margin of victory when turnout is higher (Geys, 2006) and high turnout rates favoring the minority (Bernhagen & Marsh, 1997).

A public transport bus assignment problem: parallel metaheuristics assessment

Combinatorial Optimization Problems occur in a wide variety of contexts and generally are NP-hard problems. At a corporate level solving this problems is of great importance since they contribute to the optimization of operational costs. In this thesis we propose to solve the Public Transport Bus Assignment problem considering an heterogeneous fleet and line exchanges, a variant of the Multi-Depot Vehicle Scheduling Problem in which additional constraints are enforced to model a real life scenario. The number of constraints involved and the large number of variables makes impracticable solving to optimality using complete search techniques. Therefore, we explore metaheuristics, that sacrifice optimality to produce solutions in feasible time. More concretely, we focus on the development of algorithms based on a sophisticated metaheuristic, Ant-Colony Optimization (ACO), which is based on a stochastic learning mechanism. For complex problems with a considerable number of constraints, sophisticated metaheuristics may fail to produce quality solutions in a reasonable amount of time. Thus, we developed parallel shared-memory (SM) synchronous ACO algorithms, however, synchronism originates the straggler problem. Therefore, we proposed three SM asynchronous algorithms that break the original algorithm semantics and differ on the degree of concurrency allowed while manipulating the learned information. Our results show that our sequential ACO algorithms produced better solutions than a Restarts metaheuristic, the ACO algorithms were able to learn and better solutions were achieved by increasing the amount of cooperation (number of search agents). Regarding parallel algorithms, our asynchronous ACO algorithms outperformed synchronous ones in terms of speedup and solution quality, achieving speedups of 17.6x. The cooperation scheme imposed by asynchronism also achieved a better learning rate than the original one.

Specification of a partial replication protocol with TLA+

Nowadays, data available and used by companies is growing very fast creating the need to use and manage this data in the most efficient way. To this end, data is replicated overmultiple datacenters and use different replication protocols, according to their needs, like more availability or stronger consistency level. The costs associated with full data replication can be very high, and most of the times, full replication is not needed since information can be logically partitioned. Another problem, is that by using datacenters to store and process information clients become heavily dependent on them. We propose a partial replication protocol called ParTree, which replicates data to clients, and organizes clients in a hierarchy, using communication between them to propagate information. This solution addresses some of these problems, namely by supporting partial data replication and offline execution mode. Given the complexity of the protocol, the use of formal verification is crucial to ensure the protocol two correctness properties: causal consistency and preservation of data. The use of TLA+ language and tools to formally specificity and verify the proposed protocol are also described.