Future of Media Part 1 - will User Generated Content drown in its own output?

One of the prime tenets of the Web 2.0 movement has been the democratisation of content creation, leading to Time Magazine's declaration of 2006's Personality of the Year being YOU! However, two years on I have been reflecting on what this means, and if one looks at it empirically, there seem to be two major trends:

- An expansion in the copying of copyrighted works
- An expansion in the amount of cr*p content

I suspect that one begets the other.

YOU! are the problem.

Just as all Welshmen can sing, but not all should, we get to the situation where although all people can blog / twitter / etc, when it comes to many things not all should either - it merely fills up the bandwidth with noise, reducing the signal. (This, we suspect, explains the increasing use of purloined copyright content - it is higher quality and thus more likely to be consumed, but by and large costs nothing to copy and send today)

Our hypothesis is that in essence, UGC is an inefficiently structured system - it is manageable when small as its benefits (free form expression) outweigh its problems, but as it grows and takes over more and more of the output, its inefficiencies eventually prove its undoing.


Peak UGC

Being of analytical bent I was playing with a way of modelling the Future of User Generated Content and how it may turn out. In effect the "thought experiment" analysis presented here as a result implies that we will probably reach a period (soon?) of Peak UGC, to the point where the signal is lost among the noise. The next development we will probably see is vigorous filtering. The How and where is the interesting thing.

A Hypothetical Model Part 1 - all Commenters are Equal

Firstly, imagine that in this Democratised Media (the mix of mainstream, blogosphere, PR puff, and whatever other media that is Google-able - aka the Goog-o-sphere) that a communication split is such that about 1% of these (User & Other) content Generators will Create and Communicate (aka make a lot of noise), that a further 9% will Comment (aka make some noise) and that 90% will Consume - aka say nothing. ( Ratio follows the emerging rule of thumb for Social Media of 1:9:90 Creator/Commentator/Consumer)

Secondly, imagine also that on any topic you may care to name, this same 1:9:90 split applies to Knowledge - between those who Know a Lot (1%), those who Know Something (9%), and those who Know Nothing about it (90%).

We now assume the total output in the Goog-o-Sphere on any matter is distributed equally according to probability, thus it means only 1% of the Creators know anything about the given topic, but all will Create, Communicate etc.

This gives the following truth table, if we assume All Men(/Women)'s postings Are Equal, in this most Democratic of Media



But let us assume that the vast majority - the Consumers - remain a silent majority and their posts are never made. Thus, ignoring the content theoretically produced by the consumers and summing that of the Communicators and Commentors gives a total of 10% of all possible output is posted. Unfortunately, most of it is by the 90% who know nothing, and 9% by those who only Know Something. Now, if you weight the Know Somethings as being on the button 2/3 of the time, and knowing nothing 1/3 of the time, then you get a Signal to Noise Ratio of 8%

A Hypothetical Model Part 2 - factoring in the A Team

But, I hear you say, this is patently untrue. Being of noble spirit and collegiate mind, surely those who Know Nothing will indeed refrain from holding forth on this Topic du Jour. In a spirit of seasonal generosity I grant you this, though insist that - empirically - it is my observation that some - say about 1/3rd of those who Know Nothing - will weigh in anyway. Assume similarly that not all those who Know Something will also weigh in, but about 2/3rd of them will

Also, we have to factor in the bigger voices the Creators and Communicators have vs the Consumers - so, if we assume the average Consumer has 1 volume point (where this is the reach of their view), then let us assume the Commenter - via twitter, small blog or whatever has 10 x the volume, and the frenetically Communicating Creators Pumps it up a mighty 100x. But given that the Consumer mainly says nowt, we will focus on the Great Communicators and ignore Les Autres. The silent majority will stay that way!

(In other words, Media - even Democratic Media - is more accurately termed Feudal Media - a small number of extremely media rich nobility at the top, a larger group of media bourgeoisie in the middle, and the media peasantry tugging virtual forelocks on the bottom)

This amended Feudal Media truth table is shown below, adjusting the likelihood to comment in the column 2nd left and the Volume of the commentary on the row 2nd from top. We see that the Signal to Noise ratio has doubled to 16%.




(There is, by the way, a lesson to be learned about why heirarchies came about - they are potentially better information processing systems)

Essentially, what user generated content does is to take the workload from the editor, and puts it on the end user in the form of a lower signal to noise ratio that reqires the user to do the filtering and editing. There is a limit to how much time people can spend doing this, however. Attention time has a hard stop.

This, we also contend, is what a Google Search increasingly looks like today - the algorithms reward vessels making a lot of noise, but are unable to discern if those vesels are empty or not. In gfact, much SEO wizardry is essentially trying to portray empty vessels as half full. If you observe the blogosphere you will quickly understand that noisy, empty vessels may gather links as well or even better than those who are In the Know, as they are often playing more populist - aka linkbaiting - cards.

Anyway, at some point the effort involved in deciphering the signal gets too high. Simply stated, a 16% S/N ratio means only 16% of your time spent is gaining useful signal - the rest is wasted.

In other words, at low Signal/Noise ratios, we are not looking at a "River of News" - we are looking at best at a stream of clear signal mostly lost in a vast Sewer of Sh*t. Its not so much a firehose as an effluent outlet pipe.

Now I don't know the exact estimated productivity improvement Google etc has brought in real terms, but lets assume that at some reduced S/N point the average punter says "f*ck this" and goes and does something else. We would contend that 16% S/N ratio is pretty close to that point. (12.5% S/N means for every day worked a useful 1 hour is signal - that, we contend, is roughly equivalent to the world Before Search).


How Shall we Filter

By observation, the forces driving self-moderation by the Know Nothings are being outweighed by the desire to Say Something Anyway (gets googlejuice), to say nothing of the increasing intrusion of PR, commercial blogs, reflecting media (eg FriendFeed etc) and all sorts of other new entrants to the Goog-o-sphere. Thus, we contend, the Signal to Noise ratio is by definition increasingly getting worse, not better.

Hence the urgent need to filter - now the interesting thing about this truth table is it gives some clues as to how to filter. That small % of Very Noisy Know Nothings are easy to identify and a big part of the Noise, and thus culling a small group of voices produces a near doubling of the signal to 29%, as shown below:



This is more like it. Communications theory likes signal to noise ratios of 100%+ (Analog gear likes 400%+, humans are remarkably good filters so need much less, but, as we argue above, eventually the wasted effort is just too large and they will try another way).

One can argue that Social Media is part of the filtering process too, especially of the harder-to-filter remaining know nothing commenting group - but it only really works for incoming stuff, when you go and do historical searches its all there back in the cesspit.

(In)Conclusion

Clearly this is just some (hopefully Know Something) input to the discussion, but what we are showing is this:

- Left to itself and humanity, User Generated Content will kill itself by drowning in its own output.
- Democratised Content is not democratic - a small number of empty vessels create most of the noise
- The "Publish, then Edit" model is eventually unsustainable

The only way to get some form of benefit from the "river of news" is to filter out the cr*p.

- The easiest stuff to filter are the small number of very noisy empty vessels.
- After that, it gets harder to filter - moderation, algorithms, social media are all partial solutions
- It is probably easier at that point to pre-filter the Sewer of Crap - Asymmetrically Subscribe, only go to trusted, edited sites etc
- This is hard enough to do for real time media, but far harder with searched historical media searched via Google etc as it can't - today anyway - Asymmetrically Search
- But it will have to tomorrow. The hard limits to our attention and time to be productive will force this.

Funnily enough, this is good news for edited media - ie good old mainstream media and similar - because one way of "asymmetrically searching" is to not search the whole Sh*t-o-sphere, but just subsectors which have been edited for signal, then published.

Thus, we contend, the UGC "explosion" is an aberration, is probably a temporary phenomenon, and will subside again to a far smaller piece of the pie- largely because the load it puts on the average user is too high for it to be productively sustainable.

Next Installment - Part II - what does User Generated Video look like