I would say the majority of his „discourse“ revolves around antisemitic framing. His latest conspiracy myth is that Yehudah Teichtal (community Rabi of the Jewish Community in Berlin) is the real chancellor of Germany.
All political elites he talks about are either Jews or controlled by „Jewish interests“ according to him.
He has a very simplistic Us (the righteous and morally intact) vs. Them (Jews, and their allies) worldview. He projects it into almost everything he talks about.
The first: add to the posts by Hildmann a Discourse tag, for example #ethno-hildmann. With this, it is easy, both in Tulip and in the future multi-tenancy Graphryder, to induce the SSNA of that particular group of posts.
The second: add the #ethno-hildmann posts to the ethno-poprebel corpus. We have a way to reduce the bias implied by an over-representation of Hildmann’s thinking in the corpus, based on a “one person one vote” logic: an edge in the codes co-occurrence network is only counted once per informant. When we reduce the network, we filter it, so that only edges representing associations made by, say, 4 or more different people remain. This way, associations that are made by Hildmann alone are filtered out: those that remain are those that he shares with others. The technique is explained here.
@Djan@Richard For a very similar time-scope and task, we used constructed weeks (European elections campaigns), constituted randomly, with a smaller control sample of non-electoral period in DEMOS (p. 14-18). However, I agree with Mania , I personally do not think you have to be as rigid when your intention is to study already identified specific narratives and you are approching it through CDA,
The other way out might be stating a relevance-based sampling was used, because you knew you will find it there and quantification was not the guiding principle:)
I’ve got six hours of meetings at UCL tomorrow , so I won’t be able to make the first day of the hackathon but I’ll pop in on Thursday if that’s OK. Good luck!
@amelia and All. Are we meeting tomorrow (Friday, 30 April)? I was not able to do much this week or attend the event. I got my second Moderna shot on Tuesday (in itself a half a day event) that knocked me down. Today I am much better, but had to catch up on several neglected tasks. So, I have not worked on our stuff yet.
Dear all, I just entered my coding view and basically, all my codes have disappeared from my account (-creator: Jirka_Kocian)…did the dissociation happen because of the current tinkering for the purpose of the hackathon? does anyone else have this problem as well?
Hm this is weird. I wonder if it’s because they are now all under categories which I created. The original author of each code is still present, though…
Quick notes from today’s meeting, to be fleshed out later:
Use multiple instances of codes in one post now that we are working with interview data (no need to limit to 1 per post).
Use the 6 coded Polish interviews to explore some kind of proximity index for codes (discuss with @alberto) since interviews are long-form. Breaking up interviews ourselves doesn’t make sense because we want to be able to explore them as units.
Before the next meeting, make sure to review the new categories and think about whether they work for us.
Notes from Biweekly POPREBEL Ethnography Meeting 2021-05-21T13:00:00Z → 2021-05-21T14:30:00Z
1. Splitting Interviews
We decided that the most ideal data model is to post interviews as their own topics. Each interviewer question is its own post, and each response should be attached to a pseudonymised Edgeryders account as its own post. The interview then is a thread that consists of questions by the ethnographer and responses by the interviewee.
This helps us solve our issue of giant interview posts (which prevent us from coding precision) and also the conundrum of how to “artificially” separate posts (because it follows the conversation of the ethnographer and accounts for the fact that their questions will shape the next response). This also allows us to do analysis on the co-occurrences in the interview as a whole, to better map out individuals’ thoughts and feelings.
2. Splitting existing interviews.
To make existing interviews fit the above model, we will need to split them. Already coded posts @alberto suggests that we ask dan or @matthias to help split with a script (since the Polish forum has most interviews coded). And Alberto if you could specify how you see this most easily done, that’d be great, as you explained it well in the call!
from @Wojt and @Jirka_Kocian we need a list of the interviews and post IDs so that we can do this.
3. Code Categories
We are reaching a good collective system in the backend. Before next meeting, look at the backend to see what codes could be merged. We envision being able to create an analytical system from the top-level categories, categorising interviews based on the predominant code or group of codes from each category. (e.g. Emotion predominantly anger; Ideology predominantly left-wing; Actions predominantly seeking political alternatives and questioning leadership; Places predominantly Poland).
4. Intercoder reliability.
We may want to explore some quantitative measures of intercoder reliability, like these measures. Can explore further with @alberto, who is interested / probably already nerding out
4. Agenda for next time.
Next Friday we will have a 1.5 hour meeting.
3 to 3:30 Brussels: Discussing field ethnographers comparative decisions, what to hold consistent across fieldsites.
3:30 - 4:30: In-depth discussion of backend coding categories. Merging existing codes together to create a tighter schema.
(@Wojt also bug reporting in GitHub the two issues of code authorship in backend and highlight disappearances and @jitka.kralova coordinate coding training schedule for interested field ethnographers).