Transcription
[Music] [Applause] [Music] [Applause] Imagine that this weekend you and your group of friends will try to pick a TV show to watch together. It seems a straightforward process, but everyone has their own preferences. If you are like me, you would like to watch sci-fi or fantasy—The Rings of Power—but maybe others might prefer reality (Love Is Blind), a drama (Euphoria), or a thriller (Squid Game). And there may always be the small group that would like to rewatch for the 10th time Brooklyn 99. Some might be very vocal about their favorite TV shows, while others might more easily concede. And maybe not everyone gets along. So what can we say about the final outcome of these interactions?
If you and your friends engage in a back-and-forth discussion regarding whether to binge-watch a thriller or a drama, how does this iterative process shape the group's ultimate decision? And how much effort is required from your group of friends in order to reach a compromise and finally pick one option? This is an example of a community of interconnected individuals—whether friends or adversaries—with diverse beliefs, exchanging their opinions time after time in order to reach a collective decision. And there you have it: a collective decision-making system. But let's break this down. This word has multiple examples of collective decision-making. For example, in nature, one can observe several collective behaviors in animal groups, from the selection of food sources to migration mechanisms. These collective behaviors are driven by a need to reach a consensus that benefits the whole society and avoid potential risks, disadvantages, or delays due to indecision. Consensus—the idea that all the agents or individuals share the same preferences—in these examples is facilitated by cooperation between the individuals.
However, in other collective decision-making systems, such as in our society, we can imagine that this process gets more complicated due to conflicting opinions, internal rivalries, or antagonistic interactions between individuals. Typical examples that you might be familiar with include online social networks, parliamentary systems, or sport games. The set of behaviors that we can observe in this case are not usually limited to consensus or agreement. Whether we are looking at animal groups, your group of friends, or parliamentary systems in general, collective decision-making does not reflect a random choice taken by the group, but instead it is rather a complex process that should capture and reveal the way a society functions. This is where technology steps in.
In today's world, we have different tools that can help us develop or design network dynamical frameworks or models that help us understand or even predict how collective decisions get made—or fail to get made—in complex social systems. And in this case, we can think that this is not an easy task to do. We have different societal challenges—related or interdisciplinary challenges—related to learning in social group dynamics, to data collection, to uncertain systems, to validation and demonstration. On the other side, it requires the use of adequate frameworks that need to navigate a trade-off between capturing the complexity of real-world behaviors and maintaining simplicity for the theoretical analysis. Traditionally, the main core of these algorithms has been the idea that individuals don't just interact once (otherwise we will not have any social consequence and we will stay in our behaviors); instead, they interact in an iterative process, time after time, exchanging opinions, which usually is modeled through a weighted average. Intuitively, this means that the greater the strength of the relationship between a pair of individuals, the stronger the extent of the opinion influence. This is where networks also come into play.
If you think about online social networks, we can think about the connections between a pair of individuals as being cooperative (such as from friends or colleagues) or maybe antagonistic (if you think about people that don't like or trust each other). So if everyone gets along, then we have cooperation again, and then maybe we can reach a collective decision more easily, and maybe we can even reach consensus. But when these trustful relationships are not present, that means that social tension might arise in a group, which may also further complicate the process of collective decision-making. The way we can capture this social tension in a network, we can use a measure that we call frustration. Historically, researchers have been trying to find ways to describe triangles—relationships between three people that do not conform to the principle "the enemy of my enemy is my friend"—and frustration basically extends this notion at the network level, where we have interactions between more than three people. In particular, we say that a social network is frustrated if we cannot divide it into two subgroups where each individual inside a group are friends, and we have only antagonistic or rival ties between the two subgroups.
To give you an example using real-world social networks, think about a parliamentary system where we have the parliament members as the individuals, and where the ties among them are of cooperation if they belong to the same party, or of antagonism if they belong to different parties. In this case, a two-party parliament is not frustrated, but a multi-party parliament may exhibit very high frustration. In other words, a parliament where we have a lot of these rival political parties might exhibit high social tension. In a project with my previous advisor at LSE, we wanted to use this concept of frustration and our knowledge of the dynamics of networks in order to understand how collective decision-making is affected by the level of frustration in networks or social networks. To make things easy, let me use an example again in political decision-making.
So consider a European parliamentary system where it is very unusual for a single party to achieve a majority after a parliamentary election. In this case, the consequence is that the political parties need to engage in a negotiation process, and the outcome of this process is a candidate cabinet coalition that, in order to be successful, needs to win—or at least not lose—a vote of confidence in the parliament. By collecting data over, let's say, the last four decades regarding the duration of this government negotiation phase, we have observed that this duration of government negotiation phase—indicated in terms of the number of days from the election date to the date the government is sworn in—has increased, not only in countries where this has been historically common (think of Belgium, which, to my knowledge, holds the record for length of government negotiations), but also in other countries such as Germany, Italy, or even Sweden (think about 2018). So we naturally thought, "Okay, but can we use our technological tools—so our knowledge of collective decision-making dynamics over social networks—to explain this lengthy and complex process?"
So what we did—the intuition is that when we think about the negotiation process, it's not just about numbers; it's about these cooperative-antagonistic relationships between the political parties that I illustrated before—which means that at a group or microscopic level, it's about the frustration or the social tension in our parliamentary networks. So if you think about the political parties engaging in negotiation, and the vote of confidence as the collective decision that the community will take, then we obtain a collective decision-making system. Our theoretical analysis suggested a simple hypothesis: the higher the frustration, the higher the social effort required from the community in order to reach a decision, which in this case translated into the longer the duration of the government negotiation processes. By collecting data over the last 40 years of elections in 29 European countries, we indeed corroborated this hypothesis and indeed showed that the duration of the government negotiation processes correlated well with the frustration of our parliamentary networks. Moreover, we also observed how an increasing number of parties and increasing fragmentation were adding to increased frustration in the last decades, meaning increased social tension, meaning an addition to the complexity of the collective decision-making processes.
But this isn't just about politics. From innocent examples such as "What TV shows are you going to watch this weekend?" to other societal challenges that affect us—the population—collective decision-making affects a lot of critical aspects of our lives. To make a few examples or applications that you might be familiar with, think of traffic networks, think of market shares, or international relations; think about immunization for healthcare systems. With a group of researchers at KTH, we are now looking into ways to manage energy use and sustainability in smart cities, leveraging on tools from social networks, from data-driven approaches, from experimental designs, or collecting unique data sets from test beds or living labs at the KTH. So what does this mean for us? What is the potential for us? Well, thanks to technological tools, the better we understand these processes and we integrate this knowledge in our systems, the better we can design efficient systems, policies, and also influence mechanisms that can benefit our society as a whole. Thank you.