Search This Blog

Tuesday, 7 January 2014

Burning Ears Part 1. How to use basic Information Theory and Network Theory to tell if people are almost certainly talking about you, and if people are watching you and talking to other people that you know...

Arguably, we all have an intuitive sense of when someone is speaking about us. We talk about our "ears burning"  and tell our beloved friends that "yes, we were talking about you". Human sensory perception and social cognition together make for a very powerful set of tools for discerning what others think about us. These traits have evolved over a very long period to the point where we read many social cues in many forms largely subconsciously. There are other oddities in our ability to sense what is happening in our social sphere and what other people are doing.

Studies investigating such odd phenomena as the ability to detect someone watching us from behind have demonstrated that the human sensory system is attuned to pick up such indicators as barely discernible directional changes in temperature and air current, very quiet sounds (breathing, hair moving on a turning head) and even breathing. But what about our apparently uncanny ability to discern (some of us are better at this than others) when we are being talked about, but without necessarily being able to determine by whom exactly. This is usually more associated with the skill of social mindreading or philosophical mindreading. This has nothing to do with telepathy (so far as any science has revealed) and is all about our ability to read multiple types of human verbal and non-verbal signals and to combine this skill with reading social circumstances and anticipating the behaviour of others and their likely course of action with respect to us in certain types of situation. This is mostly done by developing a theory of their mind based upon our own thinking and behaviour and the responses we witness in those we are in a position to observe closely. We have also evolved to track certain social information. We are good at this up to a point. However, when there is uncertainty and complexity in the social situation, and limited information is available to us, we become far less effective. 

Sometimes people behave unpredictably and sometimes there are a large number of possible ways they could behave. Good social mindreaders are those that can anticipate more accurately in those situations where there is greater uncertainty. As with chicken sexing - it is often the case that such adepts cannot explain their ability well or at all. However, there is nothing magical or supernatural about such gifts. They are all based upon and reduce to the physical processing of physically transmitted and stored information, and the synthesis of information from different sources over time.

Much social mind-reading behaviour can be modelled with game theory, wherein a give situation involving human agents interacting is treated as a game with a set of rules for possible 'moves' and an objective to be achieved by anticipating the other party better than they anticipate you. However, that is not what I want to discuss here. Here I want to explain how the basics of quantitative information theory can help you to determine with high confidence when someone is reporting on you, or when others are talking about you. There are two basic principles that are immensely powerful. One is that any structured situation is a prospective information source. The other is that the less likely the structure of a situation - the content of it or what is happening in it - the more certain you can be when you get what seems to be a 'message' from that situation.

The best way to understand this initially is to paraphrase basic elements of the 1948 The Mathematical Theory of Communication by Claude Elwood Shannon. Shannon’s landmark paper extended work done 20 years earlier by R.V.L Hartley. Hartley’s project was to devise formulae for modelling the transmission of information between points. One of his main strategies involved eliminating subjective psychological components from considerations about information transmission. This amounted to the elimination of meaning. In other words, transmission of information became about the transmission of physical symbols. The correct physical symbol had to be reproduced or selected from a set of possible symbols based upon a physical signal configured by an encoding device at a source (‘source’ is Shannon’s term, but it fits Hartley’s model well enough.) The production of the signal at the encoder is driven by the selection of a symbol at the source. The source forms or chooses the symbol and the encoder converts it to a specific signal.

It is at this point that things get interesting with respect to social interactions and burning ears. Shannon’s project was to optimize communications/information transmission across the channel that carried the signal. This meant that he was trying to deal with two primary problems – noise in the channel caused by magnetic and other kinds of interference, and figuring out how to choose and configure/produce signals to transmit the most frequently used symbols with the least energy and in the shortest time. What we are most interested in here is the overall outcome in terms of communication, and especially what reaches a destination from a source across a channel. (See figure 1 below.) 

Figure 1.

Shannon produced two statistical (probability and frequency based) measures. One for the measurement of the likelihood of selection of certain symbol at the source, and the other measure for the likelihood that a certain symbol was produced at the source given the symbol that arrives at the destination. Remember that there is noise and interference in the channel that carries the signal, and therefore one cannot be 100% certain that the symbol that one receives at the destination is the one that was sent. However, there are certain factors that can be used to determine how likely it is that the signal has transmitted a symbol faithfully. This all comes down to certain statistical outcomes.

Now, human beings are notoriously bad at estimating probabilities for things in everyday life, and even worse at guessing the real probability of something where the situation is complex. Social interactions are complex, and every node added to a network is (in the case of social networks – a new person or group connected by a signal to the network) increases the complexity. The more connections by channels that the new node has to the rest of the group the more complex the model becomes. Since most people are not good at estimating even basic probabilities for real situations, and since we are equally bad at guessing the number of social nodes involved in the passing of a message – people tend to pass messages thinking that it is largely undetectable to others: assuming that ‘no one will ever find out’. However, if you know the right elements of Shannon’s mathematical theory of communication, and you understand network structure and algorithms – you can quite easily undermine this assumption.

To be continued…

No comments:

Post a Comment


Does information exist apart from perceiving agents (without some thinking organism to perceive it)?

Is Philosophy Relevant to Science?