Closed lmcnulty closed 1 year ago
On the top level I agree, but we differ in approach. Let me break the table you present apart a bit.
I doubt that anyone wants to search by this – it can probably be omitted.
Editors sometimes need to, otherwise, I agree there.
Regarding language
We could make this an array that contains all languages for which reports of the incident exist.
I think this should be extracted via graphql queries on the reports rather than maintaining two separate data structures that will need to be synchronized.
Regarding your source_domain comment,
Some of the "authors" are just the sites that the reports comes from, so I think this could be combined with authors.
I disagree, authors and institutions are different things. Even the same author can shift their writing voice substantially as they change institutions/editors. The cases where the authors and the institutions are the same are exceptions where the institution does not let the authors have bylines (e.g., the economist).
Regarding, submitters | I don't think many people want to search by this.
, the submitters do!
Regarding, text | I think we should just concatenate the report texts and let Algolia search those. It shouldn't show up in the UI – if people want to read about a report beyond the title and description, they'll click the result and go to the citation page.
I think you are implying that the contents found here will be central to the card presentation in the discover app. I am good with this, but I don't think this is an "either we present reports or we present incidents" question. While it looks like you are wanting to eliminate fields, how about we add the incident info from apps/incidents
to the Algolia records so they can be presented in the Algolia-derived cards? Then the rendering of the cards could flip between an incidents mode and a reports mode and we can still search across all the report text.
Part of the value of indexing the report text is that more of the vocabulary pertaining to an incident should be covered so Algolia will be able to provide much more complete indexing.
Steps
The results of searches in the Discover app correspond to reports, not incidents. Several people have reported that they found this confusing, including Janet, myself, and the subject in the think-aloud. It seems as if this design was inherited from the time when reports were the top-level entries in the database and incidents were just an id field shared among reports. This is no longer the case.
I think we should rework the Discover app to search incidents rather than reports. This is what people will naturally expect from something called an AI Incident database. We could just try to explain it better, but I think that would be swimming against the current. Thinking about what it would mean to search incidents rather than reports, lets go through the fields that currently exist in the Algolia index:
An additional bonus to using incidents instead of reports is that the index will be smaller and we'll be able to stay on Algolia's free tier longer.