liveBook · Manning

Accountability

As explained in Machine Learning System Design

Accountability is the assignment of responsibility for specific decisions, tasks, system components, or outcomes.

When the people who built an ML system move on, someone still has to maintain it. Make ownership explicit from the start rather than relying on the original builders to remain available. section 16.0

Write down each area of accountability so the team shares an unambiguous understanding. The record can be a simple assignment or a more formal RACI matrix: section 16.1

For each task or component, distinguish:

  • Responsible: the people who do the work. section 16.1
  • Accountable: one person ultimately answerable for completion, preferably with authority to approve or reject the work. section 16.1
  • Consulted: experts or stakeholders whose opinions are sought. section 16.1
  • Informed: people who need updates but do not directly contribute to the task. section 16.1

Also record how to escalate situations the assignments do not cover. Meeting notes, follow-ups, and updates can keep responsibilities visible, but excessive syncs waste time. Documentation preserves shared understanding when people leave or join. section 16.1 section 16.3

How Machine Learning System Design covers it

How deep it goes
Defines four RACI roles and recommends an escalation route, but does not prescribe the route's steps.
Examples
Contrasts a simple assignment with a multipage hierarchy and gives typical job roles for each RACI category.
What you can do after
Apply it via a RACI matrix to assign workers, one answerable approver, advisers, and update recipients.
How it connects
Connects RACI and escalation with documentation that supports shared understanding during personnel changes.

Also explained in

Compare how 12 books explain accountability →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Machine Learning System Design. The section links point to the parts of the book it draws on.