Athagora Data, AI & Technology Principles

Technology serves the method

Athagora may use AI and modern analytical technologies to support research.

Technology is a tool.

It is not an author of responsibility.

Athagora remains accountable for what it publishes and recommends.

Appropriate uses

Technology may assist with:

• organising information
• analysing datasets
• identifying patterns for further investigation
• coding qualitative responses
• transcription
• translation
• summarisation
• research administration
• drafting
• comparison of evidence
• testing alternative interpretations

Human judgement

A machine-generated output is not evidence simply because a system produced it.

Significant conclusions, recommendations and published findings remain subject to human review.

Sources

AI-generated factual claims, citations or quotations are not treated as verified without checking genuine underlying sources.

Athagora does not knowingly publish invented:

• citations
• quotations
• sources
• statistics
• datasets

Confidential information

Confidential client information or restricted research information must not be placed into inappropriate public AI systems merely for convenience.

Before using technology with confidential data, we consider:

• sensitivity
• contractual restrictions
• provider security
• retention
• training use
• access controls
• applicable data protection requirements

People are not scores

Athagora will not use technology simply to assign definitive judgements about a person's:

• character
• worth
• employability
• mental health
• intelligence
• future sporting potential

without appropriate context, evidence and human judgement.

Sensitive inference

Athagora does not deliberately infer highly sensitive personal characteristics from weak or inappropriate evidence simply because technology makes the inference technically possible.

Bias

Data and analytical systems can reproduce weaknesses in their underlying information.

Where technology materially affects research, Athagora considers:

• missing populations
• biased source data
• inappropriate proxies
• unequal error rates
• contextual limitations

Transparency

Not every routine software tool needs a public disclaimer.

However, where AI or automated analysis materially affects how an important finding was produced, Athagora provides enough transparency for a reasonable reader or client to understand the method.

Responsibility

"AI produced it" is not an acceptable explanation for an Athagora error. Athagora remains responsible for its work.