Re:Mozilla Lightbeam – Relationships between third parties a...
by davidascher on Friday October 25, 2013 @11:42AM
(#6612253)
Crawling is certainly a complementary data collection strategy, but it's harder to avoid IP-based "filter bubble" effects w/out also deploying something akin to a bot. The hope is that by using real people using real browsers we'll collect data that reflects actual-behavior-in-the-wild.
You're right that poisoning is a potential problem if/when the data ends up useful enough to warrant poisoning.
However, this doesn't seem like a good way to collect good quality crowd-sourced data. It can be easily poisoned, and there are simpler alternatives, such as crawling and analyzing the links by themselves. (I am assuming that an entity like Mozilla would have sufficient resources for that).