Applications of Data-Centric Science to Social Design

This chapter explains the goals of this book in relation to three major aspects: human behavior, data-centric science, and social design. These three issues have been independently developing in different disciplines; however, in reality, they are deeply interrelated with one another. Specifically, the understanding of human behavior should be foundational for investigating human society. Thus, insights based on both quantitative and qualitative understandings of human collective behavior are very useful when designing social systems. Concerning design, we see how to design our social systems based on a heterogeneous membership. This chapter also discusses the three aspects (human behavior, data-centric science, and social design) independently as well as the relationships among them from the perspective of participatory design.

Data analysis is essential for understanding the phenomena observed in the actual environment. Data analysis forms a workflow consisting of data acquisition, collection, visualization and quantification, and interpretation. The purpose of data analysis is to find insights into phenomena that we have attended to and make decision-makers change their behavior. Data utilization should form an improvement cycle with Check, Action, Plan, and Do (CAPD). This chapter shows a fundamental definition of data (four types of data formats, such as time series, network, spatial, and linguistic data). Several methodological frameworks for analyzing data and how to use the results obtained from data analysis are discussed.

This chapter discusses communication and data sharing among multiple stakeholders in a product design group and describes eight relevant techniques: inputs, brainstorming, ideation, matrix expansion, persona design, data journey mapping, prototyping, and review and reflection. Inputs provide additional domain knowledge obtained from sources that may include lectures, expert interviews, field investigations, and data analysis. This knowledge can be further elaborated through brainstorming. Ideation and matrix expansion contribute to idea development by linking different aspects of a few fundamental ideas. Design is the process of detailing what, when, how, where, and for whom we construct. As products become increasingly complex, the design process must accommodate interactions between heterogeneous stakeholders from different backgrounds. Data on consumer characteristics and behaviors can be linked to product satisfaction. Finally, after prototyping, review and reflection involves the exchange of outcomes with other participants. The involvement of heterogeneous participants serves to integrate diverse knowledge and skills.

This chapter presents a viewpoint of benefits of inconvenience for designing human-machine systems. Assuming that convenience means saving time and reducing effort, efficiency, and high functionality provide convenience to users. In this sense, one principle of system design is solely pursuing convenience. On the other hand, the drawbacks of efficiency have also been pointed out: excluding users, reducing their ability, and depriving the pleasure of using systems. Other than efficiency, several principles of system design have been proposed, including emotional design and UX design. As one such principle, this chapter introduces fuben-eki, which means the further benefits of a kind of inconvenience, and shows the ideation methods for designing inconvenient therefore beneficial systems.

We discuss a voting model in which voters can obtain information from a finite number of previous voters. It is the equilibrium process. There exist three groups of voters: (i) digital herders and independent voters, (ii) analog herders and independent voters, and (iii) \(\tanh \)-type herders. In the case (i), we show that the solution oscillates between the two states. A good (bad) equilibrium is where a majority of r select the correct (wrong) candidate. We show that there is no phase transition when r is finite. If the annealing schedule is adequately slow from finite r to infinite r, the voting rate converges only to the good equilibrium. In case (ii), the state of reference votes is equivalent to that of Kirman’s ant colony model, and it follows beta-binomial distribution. In case (iii), we show that the model is equivalent to the finite-size kinetic Ising model. If the voters are rational, a simple herding experiment of information cascade is conducted. Information cascade results from the quenching of the kinetic Ising model. As case (i) is the limit of case (iii) when \(\tanh \) function becomes a step function, the phase transition can be observed in infinite-size limit. We can confirm that there is no phase transition when the reference number r is finite. This chapter is based on Hisakado and Mori (Physica A 417:63–75, 2015).

In this chapter, we discuss a voting model on networks: random graph and Barabási-Albert(BA) model. It represents how voters are given the public perceptions. A voting model is constructed by two types of voters – herders and independents – and two candidates. The voting of independents is based on their fundamental values; on the other hand, the voting of herders is based on the number of previous votes. The herders vote for the majority candidates. The information of previous votes are obtained by networks. We discussed the difference of phases which depend on geometry. We can identify two kinds of phase transitions. One is an information cascade transition which is similar to a phase transition seen in Ising model. The other is the super-normal transition. It is the transition of convergence speed. In BA model, the critical point of information cascade transition is the same as the model in the random network model. On the other hand, the critical point of the super-normal transition disappears. In conclusion, the influence of networks can be seen in the convergence speed only and cannot be seen in the information cascade transition. In this mean, we can conclude that the influence of hubs is limited. This chapter is based on the conclusion of Hisakado and Mori (Physica A 450:570–584, 2016).

In this chapter, we discuss a voting model with two candidates, C ₁ and C ₂. We set two types of voters – herders and independents. The voting of independent voters is based on their fundamental values; on the other hand, the voting of herders is based on the number of votes. Herders always select the majority of the previous r votes, which is visible to them. We call them digital herders. As the fraction of herders increases, the model features a phase transition beyond which a state where most voters make the correct choice coexists with one where most of them are wrong. Here we obtain the exact solutions of the model. The main contents of this chapter are based on Hisakado (J Phys Soc Jpn 87(2):024002-2419, 2018).

When individuals with private information make sequential decisions after having observed the actions of those ahead of them, the actions of early individuals can influence the behaviors of later individuals. Information cascade is a phenomenon where later individuals follow the majority’s behaviors of the early individuals without regard to their own private information. As the first individual’s choice greatly affects the majority’s choice, it can propagate along the subjects sequence. The question is when and how the domino effect propagates forever. In this chapter, we study several simple models of information cascade and address the question. The memory length in the sequential decisions and the number of stable state (equilibrium) play the key role. Understanding the results can help readers to understand the micro-macro features of the information cascade experiments and the betting behaviors in a horse race betting market in later chapters.

Information cascade experiment has long history. In the canonical setting of the experiment, urn choice quiz is used where a correct urn is chosen at random from two types of urns and subjects answer which urn is the correct one with private signal and the observation of previous subjects’ choices. In this chapter, we use general knowledge two-choice quiz in the experiment. The subjects answer sequentially after observing the summary statistics of the number of subjects who have chosen each option. We estimate the response function f(z) of the subjects that describe the probability of the correct choice under the influence of the ratio of correct answers of the previous subjects. As the difficulty of the question changes, f(z) changes, which results in the change in the number of equilibrium of the nonlinear Pólya urn. When the question is easy and there is one equilibrium, the domino effect disappears, and the majority choice always converges to the correct answer. When the question is difficult, there appear two equilibria, and the domino effect continues forever. If the first subject chooses a wrong option, it continues to affect the later subjects. The probability that the majority choice converges to a wrong option increases by the first subject’s choice.

Canonical setup of information cascade experiment uses two-choice urn quiz. Based on the model by Bikhchandani et al. (J Polit Econ 100:992–1026, 1992), L.R. Anderson and C.A. Holt performed an experiment where six subjects answered two-choice quiz one by one after observing the previous subjects’ answers. They observed information cascade where a subject discards one’s private information and follows the majority choice. After the experiment, many experiments have been performed, and the length T of the subject sequence reached 40 in the experiment by Goeree et al. (Rev Econ Stud 74:733–762, 2007). In the experiments, there are two urns which contain red and blue balls in different compositions. One urn is chosen as an answer at the beginning of the experiment, and subjects answer which urn is the correct one with the private signal and the observation of the previous subjects’ choices. By the control of the composition of red and blue balls in the urn, it is easy to control the precision of private signal or the difficulty of the quiz. Similar to the experiment of the general knowledge quiz in the previous chapter, as the difficulty of the quiz increases, the response function f(z) changes, which results in the change in the number of stable state. We study the correlation function of the domino effect in information cascade and understand the micro-macro feature of the system.

We consider a voting experiment using two-choice questions. An urn X is chosen at random from two urns A or B, which contain red and blue balls in different configurations. Subjects sequentially guess whether X is A or B by using information about the prior subjects’ choices and the color of a ball randomly drawn from X. The color tells the subject which is X with probability q. We describe the sequential voting process by a stochastic differential equation. The model suggests the possibility of a phase transition when q changes. When there is not the phase transition, in the limit t →∞, we can choose the correct pod. When there is the phase transition, the votes sometimes converge to the wrong equilibrium. We consider the method to estimate the ratio of red and blue balls, q using the Bayes formula, and study whether we correct the wrong decisions even if there is the phase transition.

Racetrack betting market is famous for its efficiency. The winning probability of a win betting is equal with its vote share, and the discrepancy is negligibly small. Furthermore, the accuracy of the predictions of the market participants is remarkable. Nowadays machine learning has developed much; it cannot exceed the predictions of the markets. In this chapter, we review the accuracy and efficiency of the market using JRA (Japan Racing Association) 1986–2008 win betting data. Then we study the time series data of the betting in 2008 JRA win betting market. We study how the efficiency and the accuracy improve as betting proceeds. We derive the response function of the betters and interpret it as the combination of arbitrager, independent (noisy) voter and herder.

Agriculture is characterized by few large multinationals providing machinery and others marketing seeds and treatments, but the actual cultivation is in the hands of a multitude of farmers with varying degree of sophistication. Therefore, different from other industries where standardization of manufacturing processes enables for efficiency and good quality, agriculture is ripe for improvement in this regard. Concerning food safety, the interest for smart and portable biosensor is growing, as farmers need to understand better their operations and ensure quality. At the same time, current commercial solutions are not comparable to lab results normally available only to those large enterprises mentioned above. In this article we review how an inexpensive portable biosensor called EliChip can enable farmers to perform immunoenzymatic assays and with a wider availability of important data improve the sustainability of agricultural practices currently in use. The use of this biosensor device (lab-on-a-chip disposable card, LOC) is a promising tool devoted to the detection of contaminants during the whole food supply chain with the accuracy and precision of laboratory methods. EliChip probes are based on immunoenzymatic reaction using affinity biomolecules.

This chapter investigates the quotation/transaction activities of foreign exchange markets as an observable example of man–machine collective human behavior and examines its states based on the 1-s resolution data for the period between June 2007 and June 2011. To describe the characteristics of trading activities, a simple multivariate Poisson model for both the quotation and transaction activities of the foreign exchange market is proposed. Further, a method to calibrate model parameters from actual observations is discussed. It is concluded that a fluctuation coefficient of the common mode may be a summary index assessing the collective behavior of market participants.

This chapter proposes a method to quantify the structure of a bipartite graph using a network entropy per link. The network entropy of a bipartite graph with random links is calculated both numerically and theoretically. As an application of the proposed method to analyze collective behavior, the affairs in which participants quote and trade in the foreign exchange market are quantified. The network entropy per link is found to correspond to the macroeconomic situation. A finite mixture of Gumbel distributions is used to fit the empirical distribution for the minimum values of network entropy per link in each week. The mixture of Gumbel distributions with parameter estimates by segmentation procedure is verified by the Kolmogorov–Smirnov test. The finite mixture of Gumbel distributions that extrapolate the empirical probability of extreme events has explanatory power at a statistically significant level. This method is applicable to detecting extreme synchrony in various types of socioeconomic systems.