HN Explorer

Should the web be personalized? (blog.pubexchange.com)

12 points by eappleby ↗ HN
The average person visits 2,646 web sites per month across 89 domains. That is 89 different voices that give the average person a cross-section of new and interesting topics that they may want to explore. What if all 89 voices said the same thing?

9 comments

[ 2.2 ms ] story [ 38.7 ms ] thread
lubujackson
What stuck out to me is that the average user visits only 89 different domains a month. No matter how big the web becomes we only are aware of a miniscule fraction of it.
eappleby
It would be interesting to know how many domains a user visits per day. My guess is that it would be somewhere in the vicinity of 40 which would represent the sites that user bestows the most trust to.
user24
It's an important question to ask.

When I read "What if all 89 voices said the same thing?" I thought I understood the thrust of the article; that overpersonalisation creates an echo chamber devoid of the dissent and reflective views that allow people to make their own minds up and that therefore overpersonalization has insidious potential.

I was dismayed to find that "What if all 89 voices said the same thing?" was seemingly being asked as if the answer was obviously "Oh God yes where do I sign up" (and I do understand that their target audience would likely reply as such).

eappleby
It is up to your own interpretation, but my intention was more the former. If all 89 voices say the same thing, then not only do those voices lose some credibility and a key opportunity to share their opinion, but the users lose the diversity of opinion that they were after.

I highly doubt that the 89 domains that people visit in a month each cover different topics. I'd bet that there is a high cross-over and the user is visiting them to read about what they have to say on a topic.

PavlovsCat
No, I'd rather make up my own mind, and find my own ways to find the things I like.

I also like knowing what is going on in the world in general, not just in the niche of what I already know and am. What a concept! And for example "I wonder what is #1 on [search engine] for [random word]" is a question that cannot be answered meaningfully in a fully personalized web.

I imagine it would be a huge boon for the psyops crowd (includes marketing if you ask me) though: what better setup to fuck with people, than when no two people ever see the same? And skinner boxes also are much more effective when you don't see what is happening in other boxes I'm sure.

lovemenot
There seems to be an assumption that personalisation could only ever be as good as it is now. Can't we imagine an implementation of personalisation as good or better than a site's editor could reasonably deliver to their readers in nearly all cases. Better, by the same criteria as the article advocates: diversity; richness; editorial voice etc.

Now, stepping backward from that ideal to the present, what steps could we take to reach that goal?

eappleby
That is definitely the holy grail. Personalization + editorial voice. You would think that an implementation like what you've described would require a site editor to make multiple suggestions and then there would be some algorithm that applied the correct suggestion depending on the preference of the user.

From my experience, site editors don't want to compromise on their opinions, nor spend the time it would take to make multiple suggestions. It's a tough nut. Would love to know if any companies out there are addressing that problem and what the response has been from publishers.

lovemenot
I don't know of any existing implementation, but I'd like to propose an alternative architecture.

Assume that users (readers) have access to much larger streams of content, from heterogeneous sources, than they have attention for. And that each user's stream is unique in content and may overlap with those of other users differently for every other user. So the editor of a personalised web site provides not multiple alternative content but a single filter which can be applied to each user's stream to surface related content from that user's own stream(s).