Skip to content
Humanitarian AIDJAIRI · People first
EN

First way in

Understand the impact

The same tool can help in one place and harm in another. These examples show where technology widens our understanding and where judgement must stay with a person who knows the context.

Examples from education, work and public services: when a tool helps and when human review is needed.

Traffic light: when can you rely on the tool?

Green: use it and check quickly

Summarising material you know, organising ideas, suggesting wording, ordering tasks that affect no one's rights.

Amber: use it with human review

Information you will rely on or publish, an assessment of someone's work, or material for a sensitive audience such as children.

Red: a responsible person decides

Decisions that affect a person's right to a service, a job, education or support, rely on personal or health data, or are hard to reverse.

Examples from three fields

Education

Where the tool helps
Suggesting graded practice questions, explaining a concept in different ways, helping teachers vary activities.
Where human review is needed
Assessing a student's understanding, judging academic integrity, adapting teaching to each learner's needs.
Stop sign
A tool alone deciding a student's grade or labelling them “weak” without their teacher reviewing the work.

Work

Where the tool helps
Organising correspondence, drafting, summarising meetings, analysing non-personal data.
Where human review is needed
Hiring, promotion and performance review, and any decision that changes someone's livelihood.
Stop sign
An automated system rejecting a job applicant because of a gap in their CV or their address, without a person considering their circumstances.

Public services

Where the tool helps
Directing people to the right service, answering frequent questions, sorting files by completeness.
Where human review is needed
Granting or refusing support, estimating “risk”, and any decision affecting people's right to an essential service.
Stop sign
A citizen receiving a decision with no understandable reason and no clear way to request human review.

Before relying on any tool, ask

  1. Who is affected by the result, and do they know a tool took part?
  2. Which data were used, and are they necessary and correct?
  3. Can an authorised person change the result?
  4. What happens if the tool is wrong, and who bears the cost?

Try the idea yourself

In the interactive exercise you review a decision ranked by an automated system and discover what it missed.

Go to the exercise →

From the DJAIRI ecosystem

Platforms that continue what you start here.

DJAIRI portal (opens another site)