How trust and distrust inform our choices

Every day we are faced with a deluge of information, and not all of it is mirrored in reality. How do we decide to accept something as the truth? In Applied Network Science, Giuseppe Primiero and colleagues tackle the dissemination of contradictory information.

Pixabay, CC0 Public Domain

Guest post by Giuseppe Primiero

Binary choices, in the “YES” or “NO” form, are notoriously inappropriate to answer complex questions. Nonetheless, they are used in several important situations, like political referendums (think of Brexit), or in elections which reduce to a two-contenders choice (like US Presidential elections, where a vote for candidate A is, de facto, a vote against candidate B, and the impact of any other candidate is negligible).

If their binary nature is simplistic, the way agents reach their decisions is never simple: decision-makers can be influenced by the perceived reliability of their sources and by the current distribution of agreement and disagreement on the matter under judgement.

In this sense, binary knowledge (“fact 1 is true”, “fact 1 is false”) is at the basis of information transmission in complex systems, e.g. those created by humans. In such contexts, though, agents evaluate information against previously acquired knowledge, with which it can be consistent or in contradiction. And in doing so, humans may apply certain pre-given attitudes: some agents will be sceptic, willing to assess the truthfulness (or likelihood) of the information received; other agents will be lazy, ready to simply accept information which is consistent or comes from a reliable source, and to reject it if contradictory.

As a result, trusted information will influence future interactions between the agents involved in the current exchange; and receiving contradictory information may induce to distrust future relations with the source.

We have designed a scenario that precisely characterises the above description. Our approach models binary choices on contradictory information introduced in a network by two agents (the so-called seeds).

In particular our aims were:

  1. To understand how the structure of the network affects the results of the transmission: how important is it to know the topology of the network to anticipate whether certain information will be accepted or refuted?
  2. To determine the influence of seeds on the result of the transmission: is it important if they are sceptic or lazy for the final result of the transmission?
  3. To quantify the costs of trust and distrust operations: how many times is a piece of information accepted or rejected during a transmission, and how does this relate to the final knowledge distribution of the two items?

We have created a formal model and then simulated it over four different types of networks: total (every agent is connected to any other agent), random (the links are distributed randomly), linear (the connection is directed from agent to agent) and small-world (where agents with more connections are more likely to increase their future connections with new agents, a mechanism common for example in social networks).

Our findings show that total networks reach consensus the most often, and that small-world networks are faster to reach consensus than linear or random networks. In networks with memory (where the result of evaluating the transmitted information influences trust for the future rounds), consensus is maximized if the majority of agents is sceptic. On the contrary, a network where a majority of lazy agents is present, cheaper costs for trust are shown. Most importantly, we have been able to prove that the presence of contradictory information itself is leading to more distrust.

As our society becomes more networked and phenomena of contradictory information explode (like in the recent ‘fake news’ hype), there will be a growing need for precise characterizations of the way we accept, trust and propagate data.

Read the full article here.

 


Giuseppe Primiero is Senior Lecturer in Computing Science at the Department of Computer Science, Middlesex University London (UK) since 2014. His research interests are in several areas, primarily Logic and Computation, Philosophy of Computing and Information, Agent-based modelling and Computer Simulations, and History of Computing.

View the latest posts on the SpringerOpen blog homepage

Comments

By commenting, you’re agreeing to follow our community guidelines.

Your email address will not be published. Required fields are marked *