Personalisation – Reinforcing a narrow perspective One concern with

search personalisation and similar tools is that [personalisation narrows perspectives on results]1, falling into the trap of confirmation bias – users are 1) more likely to search for affirming content and 2) if search engines customise to users, they are also more likely to return affirming results, thus users are less aware of the diversity of opinions, sources, etc. Generalist responses, personal recommendation, and diversifying * One solution to this is not to personalise results. However, this is problematic; [personalisation may be the act of a good informant]2, – we expect information to be in accord with our prior understandings, our context (geographic if nothing else), etc. * Another solution is to use friends and other social contacts as informants (see my discussion of [Facebook graph search as an epistemic tool]3) – our friends understand our common knowledge and can address this, and be interrogated as to their reasons more directly; of course, there are still biases here, and my friends may not be able to inform me about a rather large range of topics. * A solution I’ve come across today is to deliberately show results which our outside the area of enquiry, either topic wise, social wise, or perspective wise.  This is something akin to asking a friend for advice, and them giving you it and then saying “but of course, some people think that xyz”. Now, this is interesting as an approach to attempt to diversify perspectives and contexts (and certainly, as a way to diversify a results set, and avoid the “split ranking” issue raised by Simpson in which the top 50% of results are about one perspective and the bottom another).  However, it’s not  quite so transparently useful as it might seem insofar as some psychological research indicates that exposure to opposing perspectives can reinforce people’s own viewpoint (and prepare them for arguing against opposition).

XKCD testimonial knowledge

Of course, sometimes it makes sense to take (implicit or explicit) testimony as true

Diversity Aware Search – Tools The reason I’m thinking about this

today is I’ve just come across the: Diversity-Aware Scholar Search (DASS). Search engine that researchers can use to identify papers that are related to one of their working papers. DASS is user-sensitive: it down-ranks papers with authors with whom the user has co-authored papers in the past. Works on Mac and Linux platforms.  This is a cool way to explore other groups (avoid homophily in your ‘social network’/academic network) to some extent, it would be interesting to explore the same idea across domain boundaries (which I think that tool can do). And of course, as I should have known there is a reasonable amount (~[223 google scholar hits]4) of research on ‘diversity aware search’ so, something to explore further!  Ideas, tools, comments welcome here or [@sjgknight]5 [UPDATE] – this paper “Data mining, interactive semantic structuring,and collaboration: A diversity-aware method for sense-making in search” is a brilliant description of how this DAS approach might work. Verbeke Mathias, Bettina Berendt, & Siegfried Nijssen (2009). Data mining, interactive semantic structuring,and collaboration: A diversity-aware method for sense-making in search First International Workshop on Living Web, , collocated with the 8th International Semantic Web Conference (ISWC-2009) Other: 6724


  1. “Evaluating recommender systems as epistemic tools”

  2. “Evaluating Google as an Epistemic Tool”

  3. “Evaluating Facebook Graph search as an epistemic tool”