Ethics Policy

These tools describe patterns. They do not measure a person's worth.

An assessment result, a decision score, a readiness band, or a risk flag is a snapshot of behaviour, preference, or circumstance at a point in time, gathered by imperfect means. It is information to think with. It is never a verdict on someone's character, intelligence, potential, or value as a human being, and it must never be used as one.

You will not use assessment results, decision data, or any output of the Service to harm, demean, label, punish, exclude, manipulate, or pass judgement on a person. This holds whether the person is a client, an employee, a candidate, a colleague, a family member, or yourself.

This is a limit on conclusions about a person's worth, not on ordinary professional practice. Deciding who to call next, preparing for a conversation, noticing that a relationship has gone quiet, or studying how an organization tends to decide are legitimate uses of what you hold. Section 7 draws that line explicitly.

• People should know when they are being assessed, by whom, and why. Do not administer an assessment to someone under a false or vague description of its purpose.
• Tell people who will see the results before they respond, and honour what you told them.
• Do not pass off inference as assessment. Scoring someone from notes, meeting transcripts, or second-hand impressions and presenting that as an assessment of the person is prohibited: it borrows the authority of an instrument the person never took and had no chance to answer. Analyzing records you legitimately hold is a different activity, governed by Section 7 rather than by this bullet.
• Participation should be as voluntary as the situation allows. Where an assessment is genuinely required, say so plainly rather than implying a choice that does not exist.

Results produced by the instruments we administer are decision support. They are not selection instruments, clinical assessments, or determinations of fitness, and must not be used as though they were.

Separately, you may file data here from a third-party instrument that is validated for selection or another regulated purpose. Storing it in the Service does not change what it is, and does not make it ours:

• Use it according to that instrument's own validation, publisher requirements, and applicable law. We neither validate nor endorse its use for any purpose, and Section 3 of the Terms of Use governs how such data comes to be here.
• Validation does not spread. Combining a validated selection instrument with our outputs, or with any derived insight, does not extend its validation to them. The combined view is not a selection instrument.
• The obligations below apply to our outputs without exception.
• Never use a result as the sole or primary basis for hiring, firing, promotion, demotion, compensation, discipline, or any other decision with a significant effect on someone's livelihood or legal position. Our Terms prohibit this, and in several jurisdictions so does the law.
• A human being must own every consequential decision and must be able to explain it in terms that do not reduce to "the tool said so".
• Do not use these tools to build a profile that follows a person across roles, engagements, or years without their knowledge.

• Do not use results, directly or as a proxy, to disadvantage anyone on the basis of race, colour, national origin, religion, sex, gender identity, sexual orientation, age, disability, pregnancy, veteran status, genetic information, or any other protected characteristic.
• Be alert to proxy discrimination. A neutral-looking score can stand in for a protected characteristic. If a pattern in your results tracks a protected group, stop and examine it before acting on it.
• Instruments are not equally valid for everyone. Language, culture, neurodivergence, disability, and unfamiliarity with the format all affect results. Interpret accordingly, and adapt or set aside an instrument when it does not fit the person in front of you.

The Service includes AI-assisted features, and you may connect third-party AI providers. Those bring specific obligations.

• AI output is probabilistic, not authoritative. It reconstructs plausible text; it does not know things. Verify anything material before you rely on it, and never pass an AI-generated claim to a client as established fact.
• Disclose material AI involvement. If AI meaningfully shaped an analysis, recommendation, or deliverable, say so. Do not present AI output as your own clinical judgement, as a human expert's opinion, or as a validated assessment result.
• Send the minimum. Do not push personal data, assessment results, or client confidential information into an AI connector beyond what the task actually requires, and not at all where you lack the consent or authority to do so.
• Assume bias is present. AI systems reflect the data they were trained on, including its skews about people, work, and worth. Treat AI-generated characterizations of a person with more scepticism than you would treat a human colleague's.
• Do not use AI to generate assessment responses, to simulate a participant, to fabricate data, or to produce a report about someone who did not actually complete the instrument.
• Accountability does not transfer. If you act on AI output, the decision and its consequences are yours.

• Collect what you need and use it for what you said. Do not repurpose data gathered for one engagement into an unrelated use.
• Share on a need-to-know basis. Results should reach the people who need them for the purpose the participant was told about, and no wider. Assessment results are not team gossip, sales collateral, or evidence in an unrelated dispute.
• Do not aggregate a person into a dossier. Combining assessments, notes, meeting records, and third-party profiles into a comprehensive picture of an individual demands a clear purpose, a lawful basis, and — in most cases — their knowledge.
• Keep access clean. Do not share logins, do not leave sessions open on shared machines, and remove access when someone leaves an engagement.
• Correct what is wrong. If a person tells you their record is inaccurate, take it seriously and fix it.

Much of the value of this platform comes from noticing patterns in records you already hold: which relationships have gone quiet, how an organization tends to decide, where a decision is likely to stall, who to talk to next and how to open that conversation. This work is legitimate and nothing in this policy restricts it. What is restricted is presenting that work as something it is not.

Derived insight is not an assessment. An assessment is something a person knowingly takes part in and receives a result from. Derived insight is a pattern computed from records — engagement history, decision outcomes, timing, stage, prior interactions. The two carry different evidential weight, and blurring them is the actual harm:

• Never present derived insight as an assessment result, attribute it to one of our instruments or anyone else's, or file it in a person's assessment record as though it had been measured.
• Say where a conclusion came from whenever that affects the weight it should carry. "Their last three decisions stalled at the same stage" and "their DI360 shows a decision-avoidance pattern" are different claims with different standing, and only one of them was answered by the person.
• Use it to shape how you work, not to settle what someone is. Prioritizing outreach, sequencing a conversation, preparing for a meeting, surfacing a neglected relationship: legitimate. Concluding that a person is untrustworthy, low-potential, or owed less respect because of a computed score: not.
• Section 3 still applies. Derived insight must never drive a decision about someone's livelihood or legal position. For that purpose it is weaker evidence than an instrument, not stronger, precisely because the person never had a chance to answer.
• Section 4 still applies, and applies harder here. Patterns learned from historical records reproduce historical bias, including who previously received attention, budget, and opportunity. A model that ranks people by who closed before will tend to surface more of the same people. Audit it for that.
• Section 6 still applies. Purpose limitation, minimization, and need-to-know govern derived insight exactly as they govern the records it is computed from.
• Be able to explain it, and be willing to. If someone asks what your system holds about them and how it is used, answer honestly. A conclusion you would be uncomfortable for that person to read is a conclusion you should not be acting on.

Analyzing decision tendencies. Studying how a person, team, or organization tends to decide — drawn from decisions they actually took part in here — is a valuable and intended use of this platform. Where you share such an analysis about an individual, share it with them rather than only about them, and be clear that it comes from their decision history rather than from an instrument.

• Do not overstate what an instrument predicts, how accurate it is, or what science stands behind it.
• Do not describe yourself as certified, accredited, or trained in something you are not, and do not imply an endorsement or affiliation you do not have.
• Do not present our instruments, frameworks, or methodology as your own work.
• Where you do not know, say you do not know.

If you hold a credential we have issued, the following apply in addition to everything above. They are conditions of the credential as well as of the platform, and are echoed in the Certification Licence.

• Stay inside your competence. Administer and interpret only the instruments you are trained and certified in, and only in contexts your training covers.
• Never interpret a single score in isolation. Interpretation requires context: the person's situation, their own account of it, and corroborating information. A number without a conversation is not an interpretation.
• The person assessed has a right to their results. Give them their results and a debrief in language they can understand. Do not run an assessment on someone and report only to a third party without telling the participant that is what will happen.
• Protect instrument integrity. Do not share items, coach anyone to a desired result, alter the instrument, or let anyone respond on another person's behalf.
• Do not practise outside your remit. These instruments are not clinical or diagnostic. If someone discloses distress, risk of harm, or a mental-health concern, respond as a person, not as a practitioner running a tool, and refer them to appropriate professional support.
• Disclose conflicts of interest before an engagement where your interests and your client's could diverge.
• Recuse yourself where you cannot be impartial about the person being assessed.

Nothing in this policy, our Terms, or any confidentiality obligation prevents you from acting where you reasonably believe a person is at risk of serious harm. Use your judgement, contact the appropriate emergency or professional services, and tell us afterwards if the Service was involved.

If you see a use of the Service that breaches this policy — including by us — tell us at ethics@aiadvisorygroup.com. Reports are taken seriously and treated confidentially so far as we are able. We will not restrict or cancel an account for raising a concern in good faith, and doing so is itself a breach of this policy.

We may investigate suspected breaches and, under Section 13 of the Terms of Use, restrict, suspend, or cancel access, features, instruments, or a credential. Where circumstances reasonably permit we will describe the concern and give an opportunity to respond; where the conduct is serious or someone is at risk, we may act immediately.

We may update this policy. Material changes are published as a new version and require acceptance before continued use. Questions: ethics@aiadvisorygroup.com.