On the edge of noise: Recommending without adding value
In layman's terms, the Internet is overflowing with information these days. This has created huge opportunities (e.g. a definitive increase in the velocity of science), as well as huge problems (e.g. the emergence of powerful detrimental teenage subcultures) for society.
4 plus ones
Shared publicly•View activity
View 3 previous comments
- My question is: are internet users really "recommending"? I find the labels to be superficial at best, manipulative at times, designed by the web designers. Furthermore, is it possible that is the internet's way of building "links" which can then be searched and tracked? In other words, can it be a "manipulated" behaviour without the user sincerely intending to recommend without adding vallue?
If the answer is yes, and I suspect partially at least, then IMHO it might be systemic as much as behavioural issue.May 22, 2013
- if somehow (I can only think of very advanced language based AI to pull this off) there were a new way to recommend content that would verify that you actually read, and added a meaningful related opinion about the content in question that you're about to share (this would be actually another great possible use of that "brain" Kurzweil's team is building at Google, a "smart social content noise filter"), do you think people would instead choose to get that unfiltered stream of recommendations over the simple and noisy status quo?
don't you think Facebook's news-stream algorithm has been taking care of this issue by looking at your social and content consumption behavior and interests (which is pissing off a lot of page administrators that can only get to a small portion of their audience)May 22, 2013
- The comments byand made me fill in a little omission in the article. Reload it. I've added the section on "unconditional listening".May 22, 2013
- Your choice of words "unconditional listening" threw me off at first, but I assume you meant "unfiltered" / "unmanaged" or something in that vein. I think I get what you meant.
Particularly I like your line of questioning around the "model". The value add may not lie in the "listening" and "recommending" part, because by "recommending" it, the user is attributing value to it. (In most cases only tto himself / herself.) But the value lies in how the "recommendation" is modeled.
And I suspect different model adds different type of "value". Thus "noise" is something that the subscription delivers but does not fit the user's definition of "value" - probably due to it being unfiltered.
So if I can "follow / subscribe", not just to a person / community / etc, but also to the tag (or "recommendation" type) of each item and if the person tags properly, then I will get less noise. If there is no tag, I should be able to specify the filter, depending on the person followed, if I want it or not. Because I like the creativity of one person better than others.
The crux will come back to the behaviour / discipline of the "recommender". But again, I probably will just tune out or unfollow those.
That addresses the "noise" factor, but we can also add quantity filter. If I get x number of items of something, then throttle it down so I don't get too many things tto read.
A bit overkill after awhile, but I have thought of wanting something like this on my G+ and Twitter.May 23, 2013
- "unconditional" refers to the fact that once you decide to follow someone, you are submitted to seeing what they post in the future regardless of what it is, hence unconditionally.
So, I agree with everything you say.
I personally wouldn't jump into implementing clever ways of filtering based on tags or anything else. I think the entire setup of how people follow each other is deeply flawed.
It is an unstable mechanism: A mechanism whose quality depends on the good will of its users (e.g. that they tag appropriately) is not a stable mechanism. This is because there are no incentives on the publisher's side to tag things properly.
I think a fundamentally new design is needed that aligns incentives on both sides (publishers and subscribers) in a natural manner :-)May 23, 2013
- Excellent article, fascinating stuff.
It's also interesting to look at the social networking aspect in the context of the invisible tumor that has grown around it: auto-recommendations.
e.g. Essentially, YouTube recommends video B after video A if enough people watch B after A. They effectively mine recommendation value out the behaviors of their users.
So YouTube captures and amplifies the conversations that happen outside of YouTube. Two videos mentioned in the same thread or article will get linked. If Alice sends Bob a video, and Bob sends one in return, Alice and Bob's viewing history is now a record of that context, absorbed into the graph.
YouTube provides substitute value, in that they do the massive-scale curation that humans can't or won't anymore. On the other hand, it means their algorithm never creates any trends itself, it only amplifies viral trends that people have already started between themselves.
The question then is, are auto-recommendations a benefit or just a timesink. I'm leaning more towards the latter... the crack pipes of the internet, be it YouTube, Twitter or Tumblr, will always offer more content than we have an appetite for. There is never a moment when you can go "we'll, I've seen it all, time to reflect for a while".
Even worse, the time honored "top 10" list is now stereotypical clickbait rather than a sign of careful curation. We've stopped drawing lines between what's relevant and what's irrelevant, and instead created a practically never-ending spiral of irrelevance to get lost in.
I think the solution may come in the form of "slow news": aggregators that are designed to actively filter out hype and drama, and set material in a context of years rather than days.Oct 13, 2013