whoever post this didn't read the last tweet in the thread, which was the point of the whole thing. "The reality is that Duck Duck Go probably isn’t personalizing. Instead, it’s having differences in results for the exact reason we explained Google might have slight differences -- the dynamic and distributed nature of search:"
Duck Duck Go did a study using incognito mode to accuse Google of tracking users even when they're not logged in.
If you use DDG's same methodology on DDG itself, then it appears that DDG is also tracking and personalizing.
Duck Duck Go says it does no tracking of users. If that’s true, then you would expect normal and private browsing to provide exactly the same results. But if you get differences, by the report’s logic, that might suggest that Duck Duck Go is tracking and personalizing results.
This either falsifies Duck Duck Go's accusations of Google, or it shows that Duck Duck Go is tracking and personalizing.
A good part of their claim against google was that some results were unique and not local. But their definition of local was based on "local domains" which would exclude content on national sites with place specific news.
I do believe that google is too extreme in tracking and in bubbling people, but their methodology was pretty flawed.
> This either falsifies Duck Duck Go's accusations of Google, or it shows that Duck Duck Go is tracking and personalizing.
I think that is a false dichotomy; it's entirely possible that the distributed search is just part of the reason that Google's search results are different, and that DDG's accusations of filtering are also true.
I don't know if this is actually the case (I didn't look in to the matter very deeply), just saying that it's possible.
I've been waiting for DuckDuckGo to get called on passive search customization based on IP, geolocation, fingerprinting and browser state (excluding cache and cookies).
I figured it would be only a matter of time before their methodology starts to slide.
Results should be identical accross WGET, CURL, and all other browsers, and remain predictable and/or deterministic for a set span of time (minutes? seconds? milliseconds?), no matter which ISP. I might understand regional differences from country to country, but country discovery should be a well-understood and surfaced attribute of your session, and only to surface legalities and obligations DDG must adhere to.
If at human time scales, one might expect significant search result fluctuations, it should be possible to discover why, without giving up secret sauce. If people paid for a change in rank, and it went live 30 seconds ago, I'm okay with that, as long as we all see the same effect at the same time, and I'm not the only one getting targeted.
If DDG targets to the below the state level, even down as far as the county level, aside from ISP trunk analysis for content delivery, that level of detail can reveal identity, no? What if I'm, the only albino dwarf in Los Angeles county? Might the nuances of my search not reveal who I was, and where I was, based on what I was querying?
I feel that distinguishing tracking and bubbling is important.
Essentially with my limited knowledge I would say that if "generic search engine" does not profile you individually and uses only the information you provide in a single session (which include accurate location via IP and external data as political preferences of your neighbourhood) it is not tracking.
Reminder that DDG released the Python scripts they say were used to analyze Google's results a few days ago. Danny Sullivan didn't use them for this tweet: https://github.com/duckduckgo/filter-bubble-study
12 comments
[ 2.1 ms ] story [ 14.5 ms ] threadIf you use DDG's same methodology on DDG itself, then it appears that DDG is also tracking and personalizing.
Duck Duck Go says it does no tracking of users. If that’s true, then you would expect normal and private browsing to provide exactly the same results. But if you get differences, by the report’s logic, that might suggest that Duck Duck Go is tracking and personalizing results.
This either falsifies Duck Duck Go's accusations of Google, or it shows that Duck Duck Go is tracking and personalizing.
Probably the former.
I do believe that google is too extreme in tracking and in bubbling people, but their methodology was pretty flawed.
I think that is a false dichotomy; it's entirely possible that the distributed search is just part of the reason that Google's search results are different, and that DDG's accusations of filtering are also true.
I don't know if this is actually the case (I didn't look in to the matter very deeply), just saying that it's possible.
I figured it would be only a matter of time before their methodology starts to slide.
Results should be identical accross WGET, CURL, and all other browsers, and remain predictable and/or deterministic for a set span of time (minutes? seconds? milliseconds?), no matter which ISP. I might understand regional differences from country to country, but country discovery should be a well-understood and surfaced attribute of your session, and only to surface legalities and obligations DDG must adhere to.
If at human time scales, one might expect significant search result fluctuations, it should be possible to discover why, without giving up secret sauce. If people paid for a change in rank, and it went live 30 seconds ago, I'm okay with that, as long as we all see the same effect at the same time, and I'm not the only one getting targeted.
If DDG targets to the below the state level, even down as far as the county level, aside from ISP trunk analysis for content delivery, that level of detail can reveal identity, no? What if I'm, the only albino dwarf in Los Angeles county? Might the nuances of my search not reveal who I was, and where I was, based on what I was querying?
Essentially with my limited knowledge I would say that if "generic search engine" does not profile you individually and uses only the information you provide in a single session (which include accurate location via IP and external data as political preferences of your neighbourhood) it is not tracking.
maybe all Firefox incognito windows machines in Finland get specific results. no tracking needed for that.