The status of Sociology as a science is regularly disputed. The sociology is repeatedly attacked for its lack of experimentation, lack of predictive power, lack of generalization, lack of accurate measurements, and so on… This often boils down to a hard fight between natural science (often coined as “hard science”), and humanities (often coined as “soft science”). The Grievance studies affair notably revealed a possible lack of scholarship and criteria in the validation of publication in disciplines related to philosophy and sociology. It is frustrating ! The ambition of sociology is to bring knowledge on human social behavior, society behavior, the interconnection between culture and everyday life decision, the implication of cognitive bias on individuals, etc. It is a pity not to have a solid bedrock of knowledge in this field ! Beyond the scientific curiosity, engineering (especially software engineering) would have a lot of benefit to withdraw from it.

Actually, if some academics are fraud, it is perfectly possible to discourse on social behaviour without telling unfounded things. It sometimes takes a lot of documentary research and some interdisciplinarity, but it may pay off. And here comes up the book I wanted to introduce you : « Fouloscopie » by  Mehdi Moussaïd. This book is an attempt to popularize a new interdisciplinarity related to study of crowd, from the atomic level (individuals) to the macroscopic level, from the physical implication to the cultural influences, independently of the species. Yes, you read it, independently of the species : this discipline brought you to study behavior of humans but also fishes or sheeps. « Fouloscopie » is actually a new french word the author coined from « foule » which means “crowd” and « -scopie », which is the french mirror of “-scopy” in English. Hence the title of this reading note : “Crowdoscopy”, the expected mirror title of the book in English. At the time these lines are written, the book has not been translated and published in English, and it is unknown if it will finally be.

More physics than you would expect

When we talk about social behavior, we naturally think of philosophers, ethnologist, behavioral economists, etc, discussing the matter. We do not expect to see a physicist coming in, putting an experimental mechanism, made out of sands and vibrating plates, and stating “You see ? This is how a crowd works”.
Yet the most numerous contributors of crowdoscopy are physicist. And physicist come with experimental device and numerical simulation already widely used in fluid mechanics. The underlying hypothesis is “In a first approach, you can consider individuals like inert particles with consistent behaving”. The behavior of individuals taken altogether can be modeled by equations very similar to some used for description of sand flows. Of course, this initial approach is an simplification. The following approach is to give each particle a certain specificity, like for example a different velocity.

This approach works quite well with crowds as of a certain density. The exact threshold is difficult to determine and depends on what we want to observe and describe. But we can state that, from a certain density of person by square meter, the persons lose most of the constitutive specificities of their individuality to become a stereotypical particle. Under this threshold, the individual is still able to adopt much less stereotypical behaviors than sand grains’.

Another threshold where the contribution of physics makes a lot of sense, is the threshold beyond we can observe turbulence of the crowd, like in fluid mechanics. The author coined the word “crowdquake” to describe this state of the crowd where pedestrians are pushing each other forming unexpected waves of pressure. These crowdquakes are prone to make deadly victims, because pedestrians of the crowd who are close to walls or at the intersection of two crowd flows, are at high risk to be crushed to death. You can find numerous terrifying stories online, like the Love Parade Disaster in Germany in 2010, or the regular incident implying crowd crush during the pilgrimage to Mecca.

Do you want to know about another physic-related threshold ? This threshold is not on crowd density, but on speed of the crowd. It was already observed in fluid mechanics that a flow when slowed down or divided in multiple flows turns out to be faster than an unique flow. This is why you empty your bathtub faster when you leaves the cap in place than when you take it off completely. The flow is slowed down at first but in the end the divided flow is faster. If you take off the cap, the particles of the flow make the plumb like clogged and the flow slows down. This is the same phenomenon which leads to add a little separator on funnel to help pouring rice in container or salt in wash machine.
The trick was also experimented with humans and sheeps. Sheeps or humans were led to go through a gate. The crowd displacement was timed. It was observed that the individuals tend to gather in circle around the gate and to clog it. Furthermore the individuals placed close to the obstacle around the gate, like a wall, undergo a dangerous pressure, prone to cause serious injury or worse. Then, another arrangement was tested. One or several obstacles were deliberately placed on the path to the gate. The sheeps or humans were once again led to the gate and the overall displacement was timed again. In the second situation, the crowd had moved faster. The gate was not clogged by the crowd. Furthermore the situation was much less dangerous for the individuals.
Next time you are walking through a station concourse or a shopping centre, if you are seeing an obstacle in front of a gate, you will know that it is for the sake of efficiency and safety.

To make it shorter : physics tells a lot about behaviour of the crowd when it comes to qualify and quantify the emerging properties of the matter. It is intellectually rich and fundamental, but not sufficient to describe the behavior of the crowd. Humans or even animals are not just particles obviously, which makes the crowd more challenging to understand than a trivial mass of sand.

Psychology and cognition in a crowd

Copy mate !

The author tells us a story : Jewish and polish, Solomon Asch fled to USA to escape Shoah. The obedience of thousands of people through Europe to participate to this massacre, had unsurprisingly been its entire life a trauma for him. Consequently, he dedicated himself to study an important aspect of this catastrophic history : the influence of crowd on individuals. Its conformity experiment is a classic of psychology. It proved that a majority of people, if put in a larger group where they are outvoted, tend to follow the group, despite the inanity of the vote. Another important experiment, is Milgram’s. In this experiment, people are put in presence of an experimenter acting a figure of authority, and encouraged to electrically shock another experimenter acting a torture victim. The experiment went up incrementally. Each shock was supposed to be more intense than previous one. Every participant paused the experiment at least once to question it. Most continued after being assured by the experimenter. More than half of participant finally administered the most powerful shock of the protocol. Good to know, but also chilling to know…

Another differential experiment showed two things. Firstly, if a person was put through a phone with an experimenter acting a malaise, far more than three quarter of individuals call the emergency services. Secondly, if a person was put through a videoconference with several experimenter, and one of them acted a malaise, while others acting like if they do not care, only thirty percent of the individuals call the emergency services.

Some lecturer interpret these experimental results as the influence of information asymmetry, taming the empathic response. But these experiments illustrate more broadly the tendency of people to copy each other, for worse, or for better !

The tendency to copy each other naturally encourages individuals to find the best possible compromise. It materializes in politics in country like Switzerland, but it also materializes with the “pedestrian highways”, which we can see forming spontaneously in crowded streets. We can also see the tendency of people to follow tracks, in the snow or in the grass, when they exist. Animals have also these behavioral biases. Fish banks act as a one-fish because every fish in the bank adapt immediately to the closest fishes in the bank. Sheeps also tend to follow each other, and an experiment was designed to show that a lonely member of the herd is enough, to lead the whole herd.

This leads to other very interesting results. For example, during a fire drill, people tend to follow each other. If a person misleads the people in the wrong way, you get a slaughter. If nobody stood to take the lead and move people out, then you get a slaughter. During the tragic knife attack in Lyon in August 2019, passers-by surrounded and held the assaillant, pushing him to release his knife and giving up. This clearly changed the outcome of the situation and avoided to have far worse bloodshed.

Blessed are the lazy gold diggers

You have two kind of personality emerging from research : followers and explorers. These last one will tend to look for new tracks and opportunities instead of following pre-known ones. The followers are the more prone to copy, other followers or explorers. In case of gold digging, these explorers will prefer to look for a new vein when followers will prefer digging the same vein as long as possible.
An experiment showed the consequences of having these two different characters in a fire drill. The individuals were ranked by time to exit the building, and each person had a financial reward consistent with their rank. If they could not exit the building in due time, they had to pay the experimenters a severe compensation. This way the incentives were consistent with a true fire situation. The explorers took the most risk to look for the exit, with the most potential gain in the end. The followers chose one person to follow and sticked to it. The followed person was naturally an explorer. If the explorer found the right path to the exit, everybody was safe, but if the explorer took too much risk to explore all possibilities, the group was at risk. Furthermore, if the explorer finally led the group to the final exit, it was possible for some followers to have their best run at the last moment to maximize their own gain.
This follower behavior would be seen as parasitic. Some authors do not hesitate to use the expression “social parasitism”. But it is far more subtle. It depends on the scale you choose to observe the phenomenon. At the individual scale, yes, followers could be parasites for the explorers who are the most at risk. Some famous examples illustrate it.

• James Marshall, a gold digger who first discovered gold in 1849 in California, did not draw the most benefit from his discovery. He was forced off his lands and died as a church mouse.
• Bill Gates, who copied very quickly the operating system CP/M created by Gary Kidall to make its own to sell it to IBM. Today we all know which one became super famous.
• And finally, the culminating anecdote : in a computer tournament, Timothy Lillicrap won the grand prize by operating a very simple strategy : copying very fast the other competitors when they got good results.

But if you step back a little and see the whole problem scope, you realize an explorer can not always handle the whole benefit of its discovery. If you realize numerical simulations with only followers or only explorers, both profile will earn a few. This situation is called “exploration-exploitation dilemma”, and what numerical simulation teach us is : you need to have mixed group of explorers and followers to maximize collectively the gain. But the collective bargain is not always the best for individuals.

Crowd as social network

“The world is small!” says the popular idiom. “It is true” says the researches in crowdoscopy.
Our social interactions can be visualized as a web where people are linked to family, friends, and acquaintances. If we could represent this whole world wide social web, could see that anybody is, at maximum, at six handshakes from anybody else on the web. Otherwisely said, you are at maximum at six handshakes of Barack Obama, Sylvester Stallone, Shinzo Abe, or even Christiano Ronaldo.
This fact was proven by several experiments. Stanley Milgram, again, made a experiment. A mail was written for a final recipient. Instead of sending the mail directly to the correct recipient, the mail was sent to another random american. The wrong recipient was then asked to resend the mail to the final recipient if they know it, or at a person they know and they could think know the final recipient. The experiment was conducted with thousands and thousands of americans, and the conclusion was : the mail which passed from hand to hand the most never outreached six forwarding. Not bad !
The australian mathematician Duncan Watts retrieved a massive database built by students in cinema, who had the hobby to track all the relationship in Hollywood little world, centered on a celebrity of the nineties : Kevin Bacon. He proceeded to a lot of data processing to discover what Stanley Milgram already discovered.
This is possible because when most of people have at maximum a hundred relationship, some people, not always famous, may have thousands and thousands of connection. These hubs are the reason why you are not so far away from a handshake with Antoine Dupont, even if the arrows of the web are not necessarily bidirectional.
But this social network is an emerging property in a lot of domain : the electrical grid in Europe ? Small world ! The URL link on the internet ? Small world ! People who had sexual intercourse together ? Small world ! And the hubs are not necessarily the pornstars.

Crowd as influence source

How suggestible is human being ? According to some serious studies, a lot. And especially one of them impressed.
The city of Framingham has been a open-world laboratory for one of the longest study in modern science history. Since 1948, the health of around 15,000 persons through 4 generations has been monitored and their social relationships were tracked. The social network drawn from this showed a small world as we could expect it. But there is a lot more to see. The people are not clustered randomly. The smokers form altogether a cluster around an unique smoker. The alcoholics form altogether a cluster around an unique alcoholic, and relatively poorly connected to the smokers’ cluster. People close from each other on the map tend to have very similar profiles. We might see it as the illustration of the saying “birds of a feather flock together”, but the data say differently. Actually the data collected decades after decades demonstrate that a social epidemic has taken place. The tendency to be overweight was strongly related to being in relation with people overweight. The probability to get fatty has been estimated to 57% in case of direct relationship to overweight third party, 20% for a second degree relationship, and 10% for a third degree relationship. The strength of the relationship between two persons is the first factor in this phenomena of social epidemic. If you have fallen out with your neighbour, it is implausible he could share his passion for Scottish whiskey with you.

This social epidemic has been a topic of concern recently, due to the wider importance taken by the social networks online in last decade. Twitter, Instagram, Snapchat, or Facebook have been pointed out for their lack of regulation in social behaviour online, especially when terrorist groups started to exploits these platforms to conduct prograpanda campaign. But, is it really important for these platforms to monitor and moderate the content shared on their networks ? Some studies suggest it, with some nuances.
Firstly, it was demonstrated that when a notification was popped up by Facebook to show that you went voting, your online connections were more prone to go voting. In the end it was estimated, a pop-up notification was able to get 340 000 more voters to the ballot boxes.
Secondly, in another cohort of study on multiple platforms, it was demonstrated that if your news feed was rigged with negative informations, you were also more likely to share a negative publication. Inversely, if your news feed was rigged with positive news, you were more likely to share positive news online.
Thirdly, it was showed up, it is possible to deduce your political beliefs, your personal tastes, your ethnicity, or your sex, simply by relying on your publications online, with an impressive 90% accuracy.
No scoop here, human has been, is, and always will be a gregarious animal. Social network influencers look forward a prosperous future.

Wisdom of crowds

This paragraph is way shorter than previous others, as the topic is wide and would need a standalone reading note. Furthermore, I have some books (like “The wisdom of crowds“, or “The myth of the rational voter“) on my shelf, touching the same topic, and I will certainly publish reading notes related to them in the future. But the author points out another cohort of scientific studies as well as a collection of anecdotes demonstrating that, under certain conditions, a crowd can outperforms experts in any discipline.
But this outperformance is not really related to inherent wisdom of the crowd or the persons forming it. This outperformance is certainly due to a statistical phenomena called “Bias–variance tradeoff“. To popularize it very quickly and very grossly : the more you skew the crowd to have it aim a particular target, the more the behaviour of the crowd becomes globally consistent and perhaps predictable. Within a wisely engineered frame, it is possible to turn a random crowd into a wise crowd, and draw the maximum possible benefit from it.

Conclusion : what benefits software engineering can draw from crowdoscopy ?

Social interaction are a very important component of software engineering. Actually, it is the first statement of the agile manifesto.

We are uncovering better ways of developing software by doing it and helping others do it. Through this work we have come to value:

Individuals and interactions over processes and tools ;
Working software over comprehensive documentation ;
Customer collaboration over contract negotiation ;
Responding to change over following a plan ;

That is, while there is value in the items on the right, we value the items on the left more.

Agile Manifesto

Engineering, and especially software Engineering, is not just about process engineering and extreme formalism. It is about designing a product, a software, with which a user community will interact. The user community can be uniform, constituted with always the same persona, or on contrary heterogenous and constituted with several personae with different objectives and interest, sometimes contradictory. From a certain perspective, a software can be seen as an organisation, as suggest movements like algocracy, or historical events like Cybersyn in Chile, or the book “Blockchain and the Law: The Rule of Code” (which I will certainly make a reading note from someday).
If software is an organisation facility, developers must be careful with software human-machine interactions as well as with human-human interactions, or human-machine-human interactions, and even broadly crowd-machine interactions.
Furthermore, developers are engineers, working in organisations, from team-size to corporation size. Research in computer science always concern the mathematics and technical fundamentals. Research, purely in software engineering, that is answering the question “how to lead the design of a software ?” is the poor sibling of research in computer science. What we know for sure is : software development is not only about mathematics and computer science, it is also about organisation and sociology. But sociology/crowdoscopy are therefore scientific fields of interest for developers.

Crowdoscopy can help us to answer the “how-to”, but it also helps us to answer the “what to”.
Recent history already showed what softwares developers can produce in terms of social impact. Social networks are organizations literally dealing with crowds. They have been best-sellers of digital economy and researches in crowdoscopy are of interest for them. As presented in previous sections, social networks have already been experimental platforms for researchers and allowed for interesting discoveries which allowed for sensible improvement in user experience of these platforms and to put them in compliance with public policy worldwide.
Some of the discoveries of crowdoscopy have already been directly valued in softwares, like in “Predictive Crime Software“. These softwares promise to take advantage of the variety of phenomena popularized in the book, to build a model able to indicate legal authority where and when the probability of crime occurrence is the highest, and justify police patrol deployment. A true and positive contribution to the regal domain.
The book’s author also teaches us, that serious games built around a crowd are becoming quite common. Crowd has already been put to contribution to find back a ship lost offshore. Wikipedia, as well as its technical basement, MediaWiki, is also an illustration of how far a software solution built around an open user community can bring. The most impressive application of wisdom of crowd certainly is the FoldIt game, developed by David Becker, who allowed a 57,000 persons crowd to elucidate a biochemistry problem concerning the Mason-Pfizer retrovirus, which withstood the researchers for decades.

Whether about the “how-to” or the “what-to”, developers are far away from having withdrawn all benefits from the science around crowd.