I have seen lots of turkey pictures last week and this has reminded me of an anniversary: When I saw those last time I have just started using Twitter, Google+ and Facebook.
So a review is overdue, and I also owe an update to my Time-Out from social networks this summer. (If you don’t have time to read further – the headline says it all.)
I am not at all an internet denier. Actually, I had crafted my first website in 1997 and had pseudo-blogged since 2002. I made these pages – not blogs in the technical sense, but content-wise – the subject of last year’s Website Resurrection Project.
There have been two reasons for my denial of modern interactive platforms, both are weird:
- Territory Anxiety: It made me uncomfortable to have my own site entangled with somebody else’s via comments, reshares and the like. I prefer platforms that allow me to make them mine. Facebook and Google+ require you to ‘fill in form’ and put you at the mercy of their designers.
- Always-On and Traceability: For many years my job was concerned with firefighting – an inherent feature of working with digital certificates that have their end of validity embedded cryptographically. I considered it odd if panicking clients would see me sharing geeky memes while they are waiting for my more substantial responses. Notifications by corporate online communication tools conditioned me to loath any piece of technology that tried to start a conversation via flashing pop-ups.
These two reasons haven’t been invalidated completely – I think I just care less. Social media is an ongoing experiment in communications.
I am using social media in the following way: (This is not at all advice for using social media properly, but an observation.)
- If I use a network, I want to use it actively. I don’t use anything as a sole channel for announcements, such as tweeting all new blog postings (only), and I don’t use automation. I don’t replicate all content on different networks or at least there should be enough non-overlap. Each network has its own culture, target group, style of conversation.
A detailed analysis of the unique culture of each network remains maybe subject to a future post. But I cannot resist sharing my recently started collection of articled on the characteristics of the most hated most analyzed network:
How to overcome facebook status anxiety
7 Ways to Be Insufferable on Facebook
Does Facebook CAUSE narcissism?
I became a Google+ fan, actually.
- The only ‘strategic tool’ I use is a simple text file I paste interesting URLs to – in case I stumble upon too many interesting things which would result in quite a spammy tsunamis of posts or tweets. This is in line with my life-long denial of sophisticated time-management tools and methodologies as Getting Things Done (which is less down-to-earth than it sounds). I don’t believe in the idea of getting mundane things out of your head to free up capacity for the real thing. I want to keep appointments, tasks, the really important items on the to do list, and thing to be posted in my mind.
- Using social networks must not feel like work – like having to submit your entries to the time-tracking tool. I said often that my so-called business blog, Facebook site, Google+ site can hardly be recognized as such. (Remember, I said this is not perfect marketing advice.)
- I don’ care about the alleged ideal time for posting and about posting regularly. It is all about game theory: What if everybody adhered to that grand advice that you should, say, tweet funny stuff in the afternoon or business stuff on Tuesday morning? My social media engagement is burst-like, and I think this is natural. This is maybe the most important result of my time-out experiment:
- Irregularity is key. It is human and normal. I don’t plan to take every summer off from social media. I will rather allow for breaks of arbitrary length when I feel like that.
And I have found scientific confirmation through this scientific paper: The origin of bursts and heavy tails in human dynamics by renowned researcher on network dynamics, Albert-László Barabási.
The abstract reads (highlights mine):
The dynamics of many social, technological and economic phenomena are driven by individual human actions, turning the quantitative understanding of human behaviour into a central question of modern science. Current models of human dynamics, used from risk assessment to communications, assume that human actions are randomly distributed in time and thus well approximated by Poisson processes. In contrast, there is increasing evidence that the timing of many human activities, ranging from communication to entertainment and work patterns, follow non-Poisson statistics, characterized by bursts of rapidly occurring events separated by long periods of inactivity. Here I show that the bursty nature of human behaviour is a consequence of a decision-based queuing process: when individuals execute tasks based on some perceived priority, the timing of the tasks will be heavy tailed, with most tasks being rapidly executed, whereas a few experience very long waiting times. In contrast, random or priority blind execution is well approximated by uniform inter-event statistics.
Poisson statistics is used to describe, for example, radioactive decay. I learned now that it can also be applied to traffic flow or queues of calls in a call center – basically queues handled by unbiased recipients. The probability to measure a certain time between two consecutive decays or phone calls taken decreases exponentially with time elapsed. Thus very long waiting times are extremely unlikely.
The exponential dependence is another way to view the probably familiar exponential law of decay – by finding the probability of no decay in a certain time via the percentage of not yet decayed atoms. Richard Feynman gives the derivation here for collisions of molecules in a gas.
Radioactive decay – the number of non-decayed nuclei over time for different decay rates (half-lives). This could also be read as the probability for a specific nucleus not to decay for a certain time (Wikimedia)
Thus plotting probability over measured inter-e-mail time should give you a straight line in a log-linear plot.
However, the distribution of the time interval between e-mails has empirically been determined to follow a power law which can quickly be identified by a straight line in a log-log-plot: In this case probability for a certain time interval goes approximately with 1 over the time elapsed (power of minus 1).
Power-law distribution, showing the yellow heavy or fat tail. This function goes to zero much slower than the exponential function.
A power function allows for much higher probabilities for very long waiting times (‘Fat tails’).
Such patterns were also found…
…in the timing of job submissions on a supercomputer directory listing and file transfers (FTP request) initiated by individual users, or the timing of printing jobs submitted by users were also reported to display non-Poisson features. Similar patterns emerge in economic transactions, describing the time interval distributions between individual trades in currency futures. Finally, heavy-tailed distributions characterize entertainment-related events, such as the time intervals between consecutive online games played by the same user.
We so-called knowledge workers process our task lists, e-mails, or other kinds of queued up input neither in First-In-First-Out-style (FIFO) or randomly, but we assign priorities in this way:
…high-priority tasks will be executed soon after their addition to the list, whereas low-priority items will have to wait until all higher-priority tasks are cleared, forcing them to stay on the list for considerable time intervals. Below, I show that this selection mechanism, practiced by humans on a daily basis, is the probable source of the fat tails observed in human-initiated processes.
Barabási’s model is perfectly in line with what I had observed in deadline-driven environments all the time. When your manager pings you – you will jump through any hoop presented to you, provided it has been tagged as super-urgent:
This simple model ignores the possibility that the agent occasionally selects a low-priority item for execution before all higher-priority items are done common, for example, for tasks with deadlines.
It gets even better as this model is even more suited to dealing with competing tasks – such as your manager pinging your while you ought have to respond to that urgent Facebook post, too:
Although I have illustrated the queuing process for e-mails, in general the model is better suited to capture the competition between different kinds of activities an individual is engaged in; that is, the switching between various work, entertainment and communication events. Indeed, most data sets displaying heavy-tailed inter-event times in a specific activity reflect the outcome of the competition between tasks of different nature.
Poisson processes and the resulting exponential distribution are due to the fact that events occur truly random: The number of particles emitted due to radioactive decays or the number of request served by a web server is proportional to the time interval multiplied by a constant. This constant is characteristic of the system: an average rate of decay or the average number of customers calling. Call center agents just process calls in FIFO mode.
Power-law behavior, on the other hand, is the result of assigning different priorities to tasks using a distribution function. Agents are biased.
Barabási is very cautious is stating the universal validity of the power-law. He also discusses refinements of the model, such as taking into account the size of an e-mail message and required processing time, and he emphasizes the dependence of the calculated probability on the details of the priorities of tasks. Yet, the so-called fat tails in the probabilities of task execution seem to be a universal feature irrespective of the details of the distribution function.
He has also shown that these bursty patterns are not tied to modern technology and e-mail clients: Darwin and Einstein prioritized their replies to letters in the same way that people rate their e-mails today.
Considering a normal (typically crazy) working day you may have wondered why you could model that without taking into account other things that need to be done in addition to responding to e-mail. And indeed Barabási stresses the role of different competing tasks:
Finally, heavy tails have been observed in the foraging patterns of birds as well, raising the intriguing possibility that animals also use some evolutionarily encoded priority-based queuing mechanisms to decide between competing tasks, such as caring for offspring, gathering food, or fighting off predators.
Thus we might even seem evolutionary hard-wired to process challenging tasks in this way.
I am asking myself: Is this the reason why I find automated posts on social media feel staged? Why I find very regular blogging / posting intervals artificial? Why I don’t like the advice (by social media professionals) that you need to prepare posts in advance for the time you will be on vacation? What happens next – program the automation to act in a bursty fashion?
I planned to connect my Time-Out experience with Barabási’s Bursts for a long time. But now this burst of my writing it down may finally have been triggered by this conversation on an earlier post of mine.
I enjoyed Barabási’s popular-science book Linked: How Everything Is Connected to Everything Else and What It Means for Business, Science, and Everyday Life on the dynamics of scale-free networks.
There is also a popular version related to his research on bursts: Bursts: The Hidden Patterns Behind Everything We Do, from Your E-mail to Bloody Crusades. Bursts is a fascinating book as well, and Barabási illustrates the underlying theories using very diverse examples. But you should better be interested in history in its own right and don’t read the book for the science/modelling part only. Reading Bursts for the first time, I came to similar conclusions as this reviewer. It is probably one of the books you should read more than once, re-calibrating your expectations.
Further reading: Website of Barabási’s research lab.
So-called scale-free networks. The distribution of the number of connections per node also follows a power-law. Scale-free networks are characterized by ongoing growth and ‘winner-take-all’ behavior (Wikimedia, user Keiichiro Ono)