Home: TPM Feature Wiki

The Team Process Map Feature Wiki documents each of the features implemented by the TPM team in a single location.

What are considered "features?"

The term 'feature' is ubiquitous in computation (but also used inconsistently), so we feel the need to clarify here. By 'feature,' we refer to one of two types:

1. Machine Learning Features/Basic Inputs. These are characteristics of the conversation that break an interaction down into its most basic pieces so that a computer can understand it. These features may describe an interaction in simple, objective terms: for example, how many words were used? What is the frequency of each word? What is the tf-idf weighting? These features may not map to an intuitive human understanding of a conversation, but they nonetheless describe important dimensions of the conversation.

2. Behavioral Features. These are characteristics of the group's communication that are motivated by theories or concepts in behavioral science. Such features tend to be how humans describe others' actions: for example, 'coordination,' 'assertiveness,' or 'emotional regulation.' They also tend to be more complex, and are sometimes (though not always) characterized by combining multiple basic inputs. For example, one way to define 'friendliness' might involve accounting for the polarity of each utterance, in addition to the length of the utterances.

How do we document features?

Before and during the process of implementing a feature, go to the feature template and copy/paste it into the new page. Name the feature page the same as the feature's name: The serial number of the origin paper, followed by a layperson-readable name. Then fill out the rest; you should edit as you go in order to make sure the documentation is up-to-date.

Which features should be documented?

All behavioral features should be documented. ML Features that are non-trivial (e.g., word count) should also be documented.