← Portfolio · ← Collaboration 2.0
2.7 · What would have come next
Not a wish list. This is where the design was heading, and it is the part that says what Collaboration 2.0 was actually for.
2.7.1 · Measuring it, and then scoring people with it
My plan was to modify and evolve Collaboration 2.0 against measured KPIs and a steady stream of feedback from employees. Two instruments:
- Questionnaires, with the statistics behind them — the same design as 4.3, run repeatedly rather than once. Repeating it is how you get the KPIs out of what people say, and it matters most for the ones where the text on the platform will not support a strong judgement on its own.
- Analysis of the text people produce by working on the platform. Every activity on the platform is text, which means the platform is also the instrument. This is the instrumented half that 2.6.3 says was missing.
The two instruments answer different halves of the same question. And once every exchange of ideas sits in text, given how far the reasoning of large models now reaches, asking that technology for a relative estimate of how the KPIs are moving is not an unreasonable thing to want.
The second one leads somewhere the first cannot. Evaluate the quality of what each person contributes, and you can use that as an input to employee performance scoring. Today that scoring is entirely bureaucratic, and employees are extremely unsatisfied with it.
Where LLMs come in. Even without a sophisticated implementation — plain prompting is enough to start — a model can help a great deal with fair analysis of a body of text. It can produce the reasoning and the explanation alongside the score, which is the part a bureaucratic performance review never provides. Majority voting, or aggregation across the assessments of several models, is the obvious way to keep a single judgement from carrying too much weight. In more advanced scenarios there is a further step, and it starts earlier than the scoring does.
What the whole idea needs first is a way to get the material out. Every back-and-forth in the comments, every change and update, every version of a document — all of it sits on the Confluence server, which we hosted ourselves in Docker on our own VMs rather than using the hosted service. There may be an API for that; if there is not, the answer is crawling the server. Past that I would build a vector database over the arguments and the context around them. That is a separate project and needs its own design work. But it is feasible, and the return is large — provided the platform is adopted and used in the day-to-day workflow.
Two shapes for the same idea, and they are not exclusive:
- A — a mechanism like Stack Overflow’s, where the most useful contributor gets a higher score from the users themselves. Ideally that tips into a self-organizing system.
- B — model-assisted evaluation of the quality of a person’s arguments and collaboration: the logic, the referencing, whether the conduct stays inside professional and ethical bounds.
Now imagine the company where the youngest people, the ones full of sharp ideas and ambition, find a window through which a contribution means something for them. I believed those talents were like oil reservoirs the company had no way of extracting. If passionate people can find each other on a platform like this and build arguments at the highest intellectual level available to them, decisions can hardly be made as badly as they were.
2.7.2 · Structural functionalism, and functional structuralism
I have always been an advocate of the idea that structure should emerge from the functionality of individuals. That works ideally when each person’s own interest lines up with the product, or with the direction development is taking. Once political concerns enter the picture, that alignment breaks down.
Here is the paradox. Collaboration 2.0 is itself an attempt to change the functionality — the culture — of the system. I think of culture as the motor driving individual behaviour, and of structure as the bump in the road that reshapes the pattern of driving (2.1.2). So in its early phase Collaboration 2.0 has to work in the opposite direction from my usual belief. It imposes structure in order to change function, instead of letting structure emerge from function.
I do not think that contradiction is permanent. If Collaboration 2.0 ever reaches its goal — individuals behaving in a way that is aligned with the good of the whole company — then by definition the team has reached a level of maturity where structure can finally evolve out of functionality, the way I believed it should in the first place.
I expected that shift, which is why the framework carries a closed loop of feedback and adjustment from the start. For now, what I proposed to the company was that the systems engineering team process this feedback and apply it. That is not the end state. In a genuinely self-organized system — one that can scale — the adjustment cannot stay the final decision of any one group, systems engineering included. It has to happen eventually without a designated authority making the call.
2.7.3 · The end state — a system nobody has to govern
The ultimate goal is that Collaboration 2.0 reaches a setup that can maintain its own self-organization, in the sense used in 2.1.2, and needs nobody to govern it. In an ideal scenario where that goal is achieved:
- seniority and standing become apparent automatically — the Stack Overflow model again, where nobody assigns the ranking;
- the weight of a vote or an opinion becomes clear — the way authority works in a large open-source project. Satoshi created the Bitcoin repository, but the biggest and most important decisions are now made by later main contributors. And at a point where two perspectives among the leaders could not be reconciled, a fork was made, and it became successful to some degree;
- the framework can revise itself — if Collaboration 2.0 needs a change or an update, the groups are able to reach a consensus about it.
That last point is the real test. A framework that cannot be changed by the people inside it is just another structure imposed from above, which is the thing this whole document is arguing against.
⟵ Up: Collaboration 2.0 · Previous: 2.6 · Measurement, feedback and refinement · Next: 3.1 · Status, outcome and reflection →