Measuring information consumption

I have just finished a report on information consumption. We estimate that Americans consume an average of 34 GB per person per day, and 100,000 words. This is “information to the eyeballs” — we have not tried to estimate how much of that we actually assimilate.

HMI_2009_ConsumerReport_Dec9_2009

Of course, gross averages like this hide a multitude of differences. For example, TV consumption among teenagers is lower than among old people. (Excuse me, “seniors” — since I’m getting up there myself I don’t need euphemisms.)  About half of the 34 GB come from computer game playing, but gaming is highly diverse. 80% of us play games in some form, but most of the bytes come from high-end PCs with powerful graphics cards and dedicated players. And so forth.

I’m going to put a whole page on this site dedicated to the report and discussion of it. The report is available at  http://hmi.ucsd.edu/howmuchinfo_research.php .

3-D graph of TV, computer, and other media

12 thoughts on “Measuring information consumption”

1. art2science

There are a variety of good comments on the tech sites. I’ve lifted part of one, and my response, from Slashdot. Two messages, and then my response. RB

========================
This number is entirely meaningless.

Is a phone conversation “consumed” as its transcript (a few hundred bytes) or as an audio file (a few hundred kb) or a really well sampled audio file that conveys nuance perfectly (a few Mb)?

A tweet is 140 characters, but if I were to take a screenshot of a screen with Twitter (and about 20 tweets) that could be a couple of Mb.

And much of that “data” could be compressed in a meaningful way. I spend most of my day in my cubicle staring at my monitor. Does all of the visual data that my eyes are receiving (about eight hours’ worth of grey walls and a small computer monitor’s contents) count?

===========================
Re:This number is meaningless (Score:4, Informative)
by xZgf6xHx2uhoAj9D (1160707) on Wednesday December 09, @01:22PM (#30379172)

It shouldn’t be entirely meaningless. Claude Shannon [wikipedia.org] showed that no matter how you represent something, it contains the same amount of information. If I remember right, he did a study early on that showed that each letter in English text carries, on average, about 1 bit of information (in the information theory sense of “information”). You can store it in ASCII or UCS-4 or as a JPEG and even though the different representations require different amounts of data, they all contain the same amount of information: some representations just have more redundancy than others. (Sadly it’s undecidable to determine how much information something contains; otherwise compression would be a lot easier).

Unfortunately this study seems to have ignored all of that good research and ignored the whole field of “information theory” in general. The numbers they’re using on page 8 are totally exaggerated and seem to have no basis in information theory. There’s no way a “small picture” contains 8 million bits of information, and even if it did there’s no way a person could actually appreciate all that information unless they were staring at it for hours.

==================================
Re:This number is meaningless (Score:0)
by Anonymous Coward on Friday December 11, @03:36PM (#30406536)

We used a specific definition of “information,” and based on that used the actual transmission rates when they existed, whatever they were. This is not the same as a Shannon number because nobody uses optimal compression; but it’s in the same ballpark. For example:
SDTV 4 Mbps
Cellular voice 10kbps (heavily compressed)
Landline voice 64kbps
Books 1.3kbps (based on a reading speed of 250 wpm and 2 bytes per character e.g. Unicode)
IMO these are defensible within a factor of 2 or 3 (Of course I’m the lead author, so I have to think that or I would have come up with different numbers!)

There’s no way a “small picture” contains 8 million bits of information, and even if it did there’s no way a person could actually appreciate all that information unless they were staring at it for hours.

Regarding the “small picture” you must be thinking of the Internet; I was thinking of something you shoot yourself. My camera spits out 1 MB jpg all the time. In any case, we did not use the numbers on page 8; look on page 32.

The hardest to figure is a reasonable bit rate for computer games. They are hardware dependent, but also depend on type of game, style of play, video settings, etc. Suggestions welcome; hard data is even more welcome. I’ve hired a grad student to try to pin this down a little closer.

2. But I’m still confused. Am I not receiving information by sitting outdoors and looking at the natural world; or by talking to another person? Isn’t it likely that someone in the 16th century was taking in just as much information as we are now?

• art2science

A very provocative point. For our study, we defined information as “(artificial) data reaching a person.” There are many other valid definitions of information, and the limitation to artificial sources was done to make the problem tractable. I could not figure out a rigorous way to quantify “natural” sources of information, though I have not given up. Depending on the definition, it might indeed be that our ancestors received more information!

3. Mary Wickline

It seems to me that this study should be named “How Much Bandwidth” — the amount of “information” in a book or article vs the amount in a movie or game is not really comparable this way.

4. Congratulations and thanks to Professors Bohn and Short on a very worthwhile effort. I think the discussion over methodologies and definitions is useful in itself.

Hopefully, support will continue for this work so that the approach can be further refined, and we will have some ability to track trends over time through regular updates. I thought that the earlier UC Berkeley efforts that focused on data creation and storage were valuable as well.

5. Dear professor Bohn,

thank you for this very interesting study.
I’m involved in a french research about digital increasing flows better control, and your report is precious for our work.

I have three questions:
– In 4.3.2 the HMI report says that the rise on interaction is an “overwhelming transformation” that causes some cognitive changes. “These changes may not all be good, but they will be widespread”.
This is a very interesting track for me but I stay hungry for more. Could you please explain a bit more what is meant here and eventually indicate links about this research track?
– The endnote 18 gives MS productivity links that don’t seem to work… Is there another way to get these documents?
– The HMI report tells several times that complementary researches will exist notably about information flows at work. How could I stay tuned with? Is there a RSS link available somewhere?

Best wishes and happy new year for the HMI research team,
Thierry VENIN

6. Thank you very much for the answer, the post and the links.
I work with a French CNRS laboratory on IT influence on work-related stress.
My thesis is that IT technologies (more and more mobiles) create time compression (kind of a real-time pandemia), a multitasking pressure, a confusion between private and professional environments, an unlimited quantity of information and a huge appeal to be always connected.
We are driving this research following 4 axis :
– seminars with managers which allow qualitative exchanges and experiences
– a quantitative survey (1,000 managers) with a French executive managers trade union
– a software that experiment an IT flows gate and self-organization tool, running for now for about 30 managers (our gola is to get one hundred experimenters). This allow a very qualitative work with experimenters of two kinds: a group (one half) we know and meet at work and “around” and volunteers that get it over the web (the other half).
– the creation of a non governmental organization trying to link european trade-unions (both employers and employees) on the subject of IT influence on work-related stress and collective possible control measures in corporates to help putting technologies at human service
I try to maintain a blog (sorry it is only in french for now) around this research (www.cooldone.com/blog)
So, the HMI report and it’s work environments extensions, the links with information overload are very very useful for our own work.
Thank you very much for sharing and for your reactivity!
Thierry Venin