Know what changed and who changed it

What it is

Once shared content has been modified, collaborators need to be able to answer two basic questions: what changed (which objects, at which level) and who changed it (attribution). Together these form change awareness. Its division of labour with real-time workspace awareness is clean: the latter serves "right now," the former serves "what happened while I was away." The smallest usable unit of change information is the triple "object + author + time" — lose any corner, and the collaborator either does not know where to look or does not know whom to ask.

Why it happens

The default state of asynchronous collaboration is absence: you are usually not present when the content changes, and you come back to a "result state." Without change awareness, the only way to recover the difference is to reread the whole result — for long documents and complex canvases, this full comparison costs linearly more as content grows, and almost all of that cost goes to the parts that did not change. Change information compresses the comparison from "person versus content" to "person versus difference": only the increment gets read, so cost tracks the size of the edit, not the size of the artifact. Attribution (who changed it) has a second layer of function: it aims the follow-up coordination — whom to ask when in doubt, whom to align with when in conflict, how much to trust an edit once you know who made it. This is the core reason shared content needs a different information architecture from personal content: every edit to shared content is a social event and must leave a traceable actor. Change information without attribution (an anonymous "updated") does only half the job — it tells you where to look, not how much to trust it or whom to ask.

Studying it

• Paradigms: log analysis over real collaboration repositories (document version histories, open-source code repositories) examining how change notifications are received and how people return to the work; controlled experiments manipulating the granularity of change information and the presence of attribution, measuring relocation time and omission rates in catch-up tasks.
• Variables: independent variables include change-information granularity (object-level / attribute-level / whole-document notifications), whether the author is attached, and delivery timing (immediate / digest); dependent variables include time to relocate changes after returning, the proportion of changes missed, and the number of questions asked of others.
• Use in interface research: grounding the granularity of version histories, activity feeds, and in-document change highlighting.
• Methodological cautions: the real test of change awareness lies in long-term, low-frequency return scenarios; a one-shot experiment task measures short-term memory advantage — it overvalues immediate notifications and undervalues the accumulated, searchable history.

Where it stops holding

The value of change awareness grows with absence duration: someone continuously present barely needs it (real-time awareness already covers them); the long-absent depend on it heavily. Highly active objects (hundreds of small edits a day) make any itemized change stream unreadable, and the design must fall back to the digest level (see the sibling discussion of why summaries beat itemized records). In collaboration relationships with sensitivities, per-person attribution can itself become a stressor — when every edit is publicly signed, some contributors start avoiding edits or becoming over-careful; that is the price of transparent attribution.

Applying it

• Record the minimal triple (object, author, time) automatically on every edit to a shared object, and surface it next to the object with the lightest possible touch — a badge, a list row, an inline marker — never behind an explicit visit to version history.
• Organize change information by object hierarchy rather than flattening it into a time stream: the returning user's question is "what changed in this document / this region," not "what happened in the last 48 hours."
• Attribution is visible between collaborators by default, with separately controlled visibility outside the group; for trivial edits (formatting, typos) allow aggregated attribution to cut noise.
• Verification: sample return scenarios and measure the reading and asking time spent "finding out what changed"; compare before and after adjusting change-information granularity, and check whether the proportion of missed changes drops.

Related

• Same group: V2.03.2 Change summaries beat itemized records · V2.03.3 Returning after a long absence needs a diff view
• Nearby: V3.01 Versions and history · V4.03 Notification and subscription granularity · V2.02 Workspace awareness
• Search terms: change awareness · change attribution · activity feed · version history