Instalment 2: Who We Are (Big Data and Social Physics)
Instalment 2: Who We Are
“How Social Influence Affects Financial Markets… Social Physics in Finance… Cooperation Through Local Peer Pressure…Predicting Spending Behavior… Using Mobile Phones to Predict Human Behavior… Voice Analytics…”
Go back to models from behavioral economics, which say people have two ways of thinking:
slow way of thinking, which is the one we’re conscious of;
fast way of thinking i.e. just association (people use this for almost everything in life).
Two ways of thinking hence we also have two different ways of learning associated with them.
Classic way of thinking about things, the economic way, is about individuals; in social physics way of thinking, people have this social fabric; what can we do to get people to work together?
Answer is don’t give incentives to individuals, give incentives to the connections between individuals e.g. when you became more active, you don’t make any money but your buddy does.
In our experiment, it was more than four times more effective than individual incentives.
Example: Saving energy – economic incentives for individuals had not really worked; when we gave the buddies a reward, people saved energy very efficiently – for $0.50 a week on average given to the buddies, people saved 17% of their energy usage.
Comparison with global examples using economic incentives: to achieve the same outcome needed ~$1 a week.
Shows the power of the social physics way of thinking at getting people to work together.
Unique aspect of financial markets: there are a lot of social influences.
People in the financial markets are constantly talking to others: brokers, to other people, to their buddy in the market, to the other side of the market, and set up their potential deal.
People who trade by mirroring the guru traders actually make some money; those who simply copy other people’s trades make less money; but if you trade a lot, on average you actually lose some money.
Social mechanisms really play a role in making money– making you rich or making you poor.
A market is not random; did a lot of research based on data, and our major finding is that based on the social influence, there was a lot of non-randomness, there was a lot of market behaviors
that was not explained by irrational thinking, or certain kind of events happening in the world, but by people influencing and talking with each other, and getting information from each other.
A new perspective for understanding financial dynamics.
You can predict certain future movements by looking at how people act on their invest behavior socially.
With the emergence of social networks, we now have something called wisdom of the crowd
The question is does this thing work? Can it improve the trading performance of the members?
Looking at a trading platform, as time passes and the network grows and becomes more complicated, the performance increases.
In our work we took the network and was able to formalize it into one number we call the cascade probability, and to also bound this number by what we call social friction, social conductivity, which are measures of the network itself.
See what happens if we take the information cascades and see how it influences the performance of the network:
When cascade probability is too low, we get very poor performance.
As network becomes more and more complex and dense, performance starts to increase.
When you have the same information circulating in the network over and over again, being amplified and causing a decrease in performance.
The way the information propagates is more important than the source of an information itself.
At the Media Lab, the work of the human dynamics group was around modeling human behavior using mobile phone sensors.
One of the studies had undergraduate students using mobile phones for about a year, and we understood patterns of diffusion and their behavior.
The work that really led us to start the company was around what happens to a person’s behavior when they’re symptomatic – if we look at the clinical markers for major depressive disorder i.e. people isolate themselves when they have a flare up of their depressive episodes or the clinical markers for bipolar disorder, you can actually detect the behavior changes just using their mobile phone data.
Able to use that information and identify with a reasonably high accuracy which of these people are likely to be symptomatic – solves a very fundamental problem for hospitals, in a big hospital system with tens of thousands of patients, how do you know which patients need help today or this week?
The company’s app collects a variety of data to understand and predict those behavior changes.
A lot of what you communicate when you’re in a conversation is not what you say, but how you say it.
How does this technology work? Imagine listening to a conversation between two people in a foreign language that you don’t understand; you may not understand the words or the topic of the conversation, but you still can understand a lot of what’s going on in the conversation.
A lot of our social interactions and our daily activities are done with our cellphone, and through our cellphone – can we leverage that to create a new mode of health care interaction where we can get more continuous information between office visits about how people are doing, particularly in psychological health.
You can understand engagement, distress levels, participation etc between say a patient and a call center; helps the call center for example understand what’s going on in the conversation and what they can do to make the conversation better.
Data is all digital and quantified, collected and used both in real time and as an historical record of interactions.