data collection not working properly

[jcomeauictx]

Finally, in order to be able to supply therapists and other professional users with detailed feedback on the group process after a meeting, data collection has to be correctly implemented. This should be a minor issue, since we need merely to write a time stamp, user ID, and action (button press, name display, etc.)

this is working again, isn't it? or is there more functionality desired that isn't there? I haven't dug into that part of the code yet, except just enough to get it working again.

I haven’t looked at the data being produced. This was never working correctly, however. What was working just produced a single entry for the entire session.

[jcomeauictx]

more details on this, from David:

Our plan is to collect key-press information for analysis at the completion of the group therapy session. Reports on the quality of the group interaction will be compiled by scanning the sequence of key presses. Therapists subscribe to these reports, which arrive immediately after termination of a session.

[jcomeauictx]

from David: "By “keypress,” I am referring to the form button - the button users press to indicate a desire to speak."

This will be something additional, which would be delivered to only the therapist by email or some other means. I guess it could be made available in the app at a later stage, however, it isn’t anything we have to worry about now. These detailed reports will require a paid subscription by the therapist. Integrating payments is another level of complications, which can be put off until later as long as we only deal with professional therapists, etc. If we integrate payments into an app, then we have to deal with store requirements for “in app purchases”, etc, which could complicate things.

It is difficult to say what the report would look like, since we will be asking therapists what they want during the market test. In research on family therapy, Jay Haley discovered that a fixed speaking sequence like:

Mother -> farther -> child Mother -> farther -> child Mother -> farther -> child

was indicative of a schizophrenic family. In a normal family, it is impossible to predict the next speaker. So, one possible item for a report would be the predictability of speaking order. This could be a single number between zero and one - the probability of correctly predicting the next speaker. Another possibility would be to determine which time periods in the conversation were showing these pathological patterns. A therapist would probably like to see that any such patterns decreased over a session.

Another finding was that the “warmth” of a conversation could be predicted by the equality of switching pauses: The time between when speaker A stopped and speaker B started compared to when B stopped and A started. Since we will be regulating the speaker starting times, this measure will probably not be available, but when buttons were pressed during someone else's turn might yield the same information - we would need independent judgements of warmth in the conversation to identify these new patterns. It might be worthwhile to have a question or two about this at the end of the session that participants could fill in. A web questionnaire for the therapist to complete at the end of the session might also be worthwhile. In the original experiments we had a question about how frustrated people felt in the conversations. This showed that the automatic speaker selection reduced frustration. I will have to find the old questionnaires, since we will need to be doing a bit of research as the system is being developed.

[jcomeauictx]

this ability was added in pyturn.