Contribution measurement

As a group writes a document or builds a deck, Dwixel attributes the work to the member who did it. The result is a live, per-member view of who is doing what, drawn from the work rather than from anyone’s opinion of it.

Measured from real work, not a survey

Nobody fills in a contribution form. The signal comes from the editing itself: words written, words removed, and comments, each attributed to a member as it happens. Activity of this kind can show who took part, but counts alone do not capture the value of what someone contributed or how the group worked together. ¹ Dwixel treats the record as evidence to read, not a verdict to accept.

Volume is not the measure

Counting saves rewards the wrong thing. Research on collaborative writing shows that authorship can be traced segment by segment and that text has a measurable survival time: a great deal of what gets typed does not last. ² So Dwixel counts the words a person actually wrote, not how many times they hit save.

• ·Large pasted bursts are detected and de-credited, so dropping in a block of text (or AI output) does not register as authoring it.
• ·Words are counted net of what a member deletes from their own work, so padding and then trimming gains nothing.
• ·Words removed count in a small separate lane, so a member who improves the work by cutting is not read as a non-contributor. The weighting is deliberately modest.
• ·Comments count toward contribution but are capped, so nobody can lift their share by commenting a lot.

Deterministic and explainable

There is no opaque model deciding a student’s share. The scoring rules are fixed, published and inspectable, which is the minimum for anything that touches a grade. It is not sufficient on its own: activity data captures what is easy to count, and being open about a method does not by itself make the measure valid. ³ The contribution figure is one input to a mark, never the mark itself.

The picture behind the number

Dwixel reconstructs who typed each sentence by comparing saved versions of the document, so the record is not just how much someone wrote but which parts survived, who cut whose work, and when it all happened. The same history draws a picture. Instructors who showed a visualisation of this kind to their classes reported that students found it made the invisible visible, ⁴ and it is a great deal easier to check against your own memory than a percentage is.

The axis matters more than it sounds. Spacing every save equally de-emphasises revisions that arrive in quick succession, while spacing by date is revealing of the rhythms of collaboration. ² Three weeks of silence followed by one long night is a real pattern, and it is invisible on the wrong axis.

What it does not measure

A great deal of real contribution leaves no trace in a document. Thinking, arguing a point, holding a group together when it wobbles. Organising is the one Dwixel can partly see, because tasks and milestones carry an author, so who set the group up is shown beside the share — deliberately beside it and never inside it, since a low figure there can be the assignment’s fault rather than the student’s. Students who type quickly, or who are writing in their first language, will tend to produce more words, and that is not the same as contributing more. ³ Read the figure alongside the work and alongside what the group tells you, not instead of them.

Students see their own data

Every student can see their own contribution share and exactly how it was calculated, in the same view their instructor gets. Nothing about the measurement is hidden from the people being measured, which follows the principle that a group and its instructor should both be able to see who contributed what. ⁴ It is worth being precise about how much that settles, though, because it is less than it sounds. A profile detailed enough to be faithful to what happened is too complex to take in at a glance, and one simple enough to take in is not faithful; simplicity and fidelity cannot both be had. ³ Showing someone a dashboard also places the responsibility on them for a system they did not choose and cannot decline while remaining enrolled. Openness is a floor here, not a defence.