Support conversation groups

[daveey]

Problem:

Sometimes a conversation can start out broad, and then split into sub-conversations. Currently this would require that moderators create new conversations, and ask users to re-vote on statements. It also means that group affinity is not preserved across the conversations.

Suggested solution:

I can imagine two possible solutions, but am not familiar with the code to know which one is cleaner.

1. Add support for a "conversation group". Conversations can be assigned to a group, and would share grouping state. Add the ability for the moderator to move statements to other conversations in the group. After a user responds to all the questions in one conversation, suggest another conversation in the group, or give them a list and let them pick.

2. Support "subtopics" inside a conversation. Any statement can be tagged with one (or more) subtopics. Each subtopic would have their own url, and prompt. Only statements tagged with a subtopic will be shown.

Thank you for your consideration

[metasoarous]

Hey @daveey. Thanks for bringing up this idea for discussion.

We recognize that addressing subtopics within a conversation can be very useful for analysis. However, the simplest way to deal with this is to just assign topic labels to comments, and use these labels in the analysis process.

Right now, this is only possible by exporting the data from a conversation, manually labeling comments, and performing your own custom analyses. We are tracking an issue (and would be open to PRs) for adding a topic/tag labeling interface to the moderation interface (see #343). This would simplify the labeling process, but manual analyses would still be required to take advantage of them. However, we've already done some custom analyses (which I think you can find in the https://github.com/compdemocracy/analysis repository) for taking advantage of these assignments, and I think we might consider adding some of these features to the automatic report.

There is certainly nothing currently stopping an administrator from inviting a cohort of individuals to separate conversations, and it's possible to see how those conversations relate to each other if you use our xid system to join participant identity between conversations. However, use of the xid system requires embedding the participation interface in a separate web page, unless you're hosting your own instance of Polis, in which case this information can be obtained directly from the database itself (we do not share this information though, for privacy concerns).

All that having been said, it's worth noting that if it's easy for participants to self-select which conversations (or "sub-conversations") they participate in, this could easily bias the results of those conversations. In some cases, this may not be a big deal, but it would be easy for a moderator who wasn't cognizant of this possibility to get themselves in trouble while attempting to interpret the results, and so we're likely to avoid making it easier to shoot oneself in the foot in this manner, given that it is ultimately still feasible to do this in cases where it makes particular sense for one reason or another.

One case where this does make quite a bit of sense is for longitudinal studies over time, to see how opinions around a topic (or topics) evolve. While we might eventually decide to support features like this, it would add a lot of complexity to the administrative interface, and I think we'll remain reluctant to do this given that fact that it's currently possible via the xid system.

Thanks again for asking about this, and please let us know if you have follow up questions about the above.