Personalized news recommendations and privacy are not mutually exclusive
0x65.dev
0x65.dev
Consider tags. Imagine if in your history, all the tags associated with an article were saved too. Tags aren't provided by the website? Websites should be encouraged to provide better tags for personalisation (tags over tracking beacons!), or the browser/browser plugin uses some kind of text processing (machine learning?) technique to generate tags automatically from content, and can train on the millions of already tagged articles in websites around the place.
The trivial solution is then: increment the count for each of the tags when a user visits a page with those tags. In top news, show the stuff at the top with the highest tag count. When doing a search term for something ambiguous that contains some of the tags, show the highest visisted tag results first. e.g. if you search for "election results" and the user has a higher frequency of "us politics" tag visits compared to "uk politics" tag visits, how US election results higher up. (You could weight this based on current events too, e.g. which one is more recent and so on). Over time, the plugin itself could have a model which could train its filtering, and maybe receive global anonymous updates that improve the algorithm. All data is local, the only thing you need is a list of tags for each article.
Just last week I started building a feed reader which does exactly that.
Feeds are more like entire newspapers, but maybe you're only interested in the tech articles. You can build up a weighted distribution of keywords over time which describe the current interests of the user, using the approach you described. And all you need to do to make it work is read as usual.
If you want to try it out: https://www.feedist.io It's still in beta but will very likely leave it next week.
Alex requests The Washington Post and CNN.
Bob requests The Wall Street Journal and Fox News.
And Joe requests PornHub.
I know the last one is not in the list buy my point is: the domain is more than enough to profile you for profitable reasons.
Plus, the fact that this isn't addressed or even mentioned in the article is not good.
Will we ever get a browser that dose not send anything to the company/organization developing it for any reason/excuse?
I wrote a concept with (possibly local) curators to subscribe to and the ability to up- or downvote them in your news stream. This way everything stays on a 1-to-1 basis.
There is no AI judging your clicks (quantitatively I click more cat pictures than 7-page articles, even if I prefer the latter).
And no grouping/democracy which I think does not work well for taste-stuff (everybody can agree on cute cats and your bubble agrees on political opinions that are obvious to you, but thought provoking, hard to digest stuff is out again).
If anyone is interested I can send you my write up (even has a few pictures in it!).
Like as described this seems like it could be defeated by a naive top(3, count(sites)). I guess it helps but the there’s probably more ambient noise from NAT and changing addresses than is being added by the random options.
Noise is added for what is called plausible deniability. Note that messages themselves do not contain any user-identifier. We take extra measures to strip request headers not needed by the server to avoid extra information that could be used for implicitly linking messages[1]. Possibility of linking messages based on network fingerprinting (ex: IP - which we do not log) still exists and is an open concern which we will solve in the next version. This at most makes it possible for us to learn that these 3 domains are visited by the same person, again - given that the list of domains are from shortlisted top-news domains, it is safe to assume that they do not contain any PII.
That said, it is not the strongest model we apply -- due to resource-constrains we have not updated to strongest models like we do on more sensitive data - via HumanWeb[2]. We will soon do the changes on a couple of dimensions: a) each domain as separate message, right now this introduces un-wanted spatial correlations, and b) send the domain through the proxy network HPN[3].
Ref: [1]: https://github.com/cliqz-oss/browser-core/blob/7679c40aec9fe... [2]: https://www.0x65.dev/blog/2019-12-03/human-web-collecting-da... [3]: https://www.0x65.dev/blog/2019-12-04/human-web-proxy-network...
Disclaimer: I work for Cliqz.
it's impossible to keep the promise of delivering quality content, when their sourcing for the "What you need to know / Top News" feature submits itself to the rules of a 24 hour news cycle[1]:
[1] https://en.wikipedia.org/wiki/24-hour_news_cycle#Critical_as...
... and ...
> footnote: The first version of Cliqz News recommended articles to users in the form of a web application by harnessing the power of the crowd. Topics were represented by large groups of Twitter users with influence in a particular topic. The content circulated within these groups was then classified and ranked, allowing us to make predictions about stories you might want to read based on detected similarities with other users’ interests
They realized they could predict what will become newsworthy by looking at twitter echo chambers. Twitter is awesome because we get to watch how (especially Western) news break, and how larger established outlets cover the same story, and (with some manual digging) follow discussions/reactions in different echo chambers and how these continue to affect the coverage. Twitter feels (or in the past felt) like grassroots journalism. The problems of the 24hr-News-Cycle aren't created by Twitter and already existed with tabloids. But Twitter amplifies the problems mentioned on Wikipedia tremendously, while even creating some new ones. Cliqz doesn't look like it is going to solve any of this, instead they assume nothing is wrong with how things work currently. Cliqz inability to think about this problem, while at the same time building / peddling services on top of it, is a red flag imo.
EDIT: the wikipedia page has a brilliant book[0][1] linked at the bottom ("Skyful of Lies & Black Swans") which I highly recommend:
[0] summary: https://reutersinstitute.politics.ox.ac.uk/our-research/skyf... [1] pdf https://reutersinstitute.politics.ox.ac.uk/sites/default/fil...
Cliqz News covers the “news cycle” consisting of “the media reporting on some event, followed by the media reporting on public and other reactions to the earlier reports” mentioned in the Wikipedia page you base your comment on, by the nature of the product. As you might want to read what you "need to know" are most relevant news events at any time of day. We look 24hours back to get the all news articles from hand-picked, trusted news domains in a country and follow event developments. the process is started every hour. There are no social signals involved in the development of this product.
Below the product statement again for further clarification “The product translates into a limited, per-country list of news articles, aiming to keep Cliqz users informed about current events. All articles originate from hand-picked, well-respected and trusted news outlets in a particular country. In order to determine the most relevant news, we first collect and process all articles that were published by those outlets during a defined timeframe, which we then compare among each other to form clusters around current events. The impact of these events, their prevalence on news sources, their presence on homepages and their times of publication all help to curate a list of Top News—updated hourly or every time a major story breaks.”