In the last few blogs I wrote about the problems with [cognitivist models of epistemic beliefs and a discourse-oriented approach to viewing epistemic action]1. I also elaborated on this view in the context of [discourse ‘to do’ information retrieval and IR as a lens on epistemic action]2, discussing some of the epistemic implications of particular moves in IR tasks. So I’ve been thinking about this a little more, and I wonder if we can think about epistemic action in IR through a kind of architecture metaphor…for any problem we can imagine two (or more) architects discussing where to draw in window frames, what view we’ll see, how the materials will taint the view (maybe they get this wrong sometimes, or forget to consider it) – how the windows will impact on the [information] architecture. In this analogy, the windows provide the perspective on the world, with individual panes representing queries, and the talk between collaborators ‘framing’ those windows. This gives the talk primacy, which I think is where we want to be, but shows us the limits of any particular section of talk – a set of queries (panes) with a particular view (window); a search session. This segmentation allows us to explore the language use and development throughout, and use a temporal analysis to compare panes within and across windows (and indeed, ‘houses’…). This would allow, for example, the sort of non-corpus and non-expert comparator based analysis described for example in: Introne, Joshua E., and Marcus Drescher. “Analyzing the Flow of Knowledge in Computer Mediated Teams.” In Proceedings of the 2013 Conference on Computer Supported Cooperative Work, 341–356. CSCW ’13. New York, NY, USA: ACM, 2013. doi:10.1145/2441776.2441816. I’m also looking at: Golding, Clinton. “We Made Progress: Collective Epistemic Progress in Dialogue Without Consensus.” Journal of Philosophy of Education (2013): n/a–n/a. doi:10.1111/1467-9752.12010. to explore assessment criteria for tasks such as this where it is possible not to come to a convergent outcome (i.e., a decision), the topics in discussion are dynamic and not based on a corpus definition, and we’re interested in the use of epistemic artefacts across multiple contexts…. more on all that in another post though UPDATE: 08:58 12/03/2013 – On twitter Golnaz Arastoopour says:

@sjgknight I like the window metaphor! But are these windows grounded in anything? What point of view do the windows represent?

— Golnaz Arastoopour (@HeyItsGol) March 12, 2013

So what provides the perspective, what point of view do the windows represent? Well, on the one side it’s the frames, windows and panes out onto the world which shape our experience. So that’s the ways we interact with the tools to view the (information) world, and the joint decisions come to in this respect (perhaps this co-constructed ‘bounding’ should be part of an assessment criteria, perhaps the Golding article might contribute here). In this sense our view on the world is grounded in our ability to construct that view, the purpose for which it is being constructed, the materials at our disposal. That’s why I think considering the [epistemic properties of tools such as google]3, [facebook graph search]4, [recommender systems]5, and [diversity aware search]6 are important. These things are the materials in the window panes, and influence the way we talk about framing our view on the world. On the other hand, it is the [epistemic] environment (including such tools) which shapes our view – some environments are very friendly (some are downright dull), others are rugged – this is about the complexity of domains, and the requirements of the task. Sometimes the exercise of relatively little epistemic effort is adequate for crediting an agent (see my [MA thesis]7), sometimes we expect rather more. Similarly, sometimes domains are just quite simple, other times they are full of nuance and complexity. Perhaps automatic classification may tell us whether we’re in a 2-d environment in the general sense or if is is more complex than that – [diversity aware search]6 may address this, with a key point being that in a 2 day environment we only need one set of windows, one perspective on the results. This is about why we’re framing, what we want to see, what will the window illuminate, etc. [UPDATE: 20/03/2013 18:50] slight note to self, but Analysis potential: 1) exploratory talk, it’s about being receptive to, and critical of, ideas 2) segmenting according to query, allows us to assess the topic of each segment (by query) what’s been explored, and potentially: 3) explore the epistemic vocabulary for each topic, this might allow e.g. a sentiment style grid with topics as rows and a rubric for epistemic dialogue as columns. Because we’re looking for exploratory talk & epistemic vocab (e.g. evaluation) blind disagreement or acceptance even where within a critical position won’t be accepted. Also, critical searches – e.g. looking for critical literature on a topic – should be ok…as should simple fact finding events (which are in any case short so could perhaps be removed)……something to test in any case. [UPDATE 15/04/2013 13:19] again a note to self: I wonder if the best analogy here is of the blind joiner (for windows in partic) – we investigators come in & can tell something about the panes, windows and frames – what size they are, and shape (but not  everything), and what the frame might be for, how it’s held together, its maturity, etc. But we don’t have access to what it’s looking out on (although we might be able to infer some of it – portholes are more likely to be seen at sea…)


  1. “Epistemic beliefs and games”

  2. “Epistemic Behaviour with Epistemic Tools”

  3. “Evaluating Google as an Epistemic Tool”

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

  5. “Evaluating recommender systems as epistemic tools”

  6. “Personalising for Diverse Results – Diversity Aware Search” 2

  7. “MA Thesis”