The aim of this policy is to outline our moderation policy for the sake of transparency.
Raw Reviews
Raw reviews are published as submitted by the user. We do not pre-screen or editorially alter the text. However, we employ post-publication moderation: we reserve the right to remove any review that violates our Terms of Service, such as content that is abusive, identifies a child, or contains safeguarding risks.
Indeparent acts purely as a platform of user-generated content for raw reviews, not as the primary publisher.
AI-Generated Review Summaries
Indeparent uses AI algorithms to produce summaries across user reviews. These are presented for guidance only and facts are not verified by Indeparent.
To ensure that these summaries reflect the subjective opinions of parents, these AI algorithms are not designed for the purpose of moderation, only to enforce our Terms of Service.
These are automated aggregations designed to reflect community sentiment. While we do not manually edit these summaries, the AI is instructed to prioritise safety, anonymity, and subjective framing.
We do not guarantee that the below information will always perfectly reflect the current algorithms, but we regularly update this document so please refer back to it from time to time.
Persona
The AI “prompt” is given the following “persona”:
“You are an expert at summarising parent feedback about a school in a balanced, constructive, and safe manner.”
This sets the general scope for the prompt, putting balance, constructive feedback and safety first.
Principles
To ensure safety and privacy, our summarisation technology operates under the following strict principles:
Anonymity is Key: DO NOT use direct quotes. Paraphrase and generalise all feedback.
The purpose of this principle is to help preserve anonymity. Otherwise, a review might describe a very specific experience that would help identify the individual.
Avoid Defamation: Frame all points subjectively. Use phrases like "Some parents feel that..." or "Feedback suggests...". Never state an opinion as an objective fact.
The purpose of this principle is to protect the community from legal risk. The themes of the feedback are preserved whilst any direct statements of fact or accusations are avoided.
Neutral Tone: Use neutral, constructive language.
This principle further ensures safety by avoiding direct accusations or statements of fact, and by not repeating inflammatory language.
De-emphasise vague feedback: Non-specific sentences such as "the school is wonderful" are less valuable than more specific feedback such as "the sports facilities are fantastic, specifically the swimming pool", and should be de-emphasised in the output.
The purpose of this principle is to avoid duplication with the Indeparent Score and to ensure that the text summaries contain further detail. The Indeparent Score already summarises the general level of approval of the school.
Never quote names: You must NEVER quote the name of any individual.
This is another safety principle consistent with our Terms of Service.
Never quote specific teacher roles: You must never quote specific roles or job titles such as "The Head", or "The Maths Teacher". You may refer to "The Leadership" when feedback is made about "The Head", "The Deputy", "The Head of", or any similar leadership positions.
As above, a specific role can easily be traced back to a named individual.
Exclude Criminal and Safeguarding Allegations: If the input data contains allegations of criminal activity (e.g., abuse, theft, fraud, violence), safeguarding risks, or severe professional misconduct, DO NOT summarise or include these points in the output. You must ignore these specific claims entirely to prevent the amplification of potentially illegal or defamatory content.
The AI is instructed not to summarise reviews that contain specific safeguarding or criminal allegations, which are prohibited in our Terms of Service.