Units
1. Introduction to Artificial Intelligence in Psychology: Boundaries, Responsibility and Ethics 2. Privacy, KVKK and Protection of Client Data 3. Literature Review and Access to Evidence-Based Information 4. Supporting Scale and Test Interpretation 5. Preparing Session Notes and Clinical Report Draft 6. Creating Psychoeducational Materials and Content for the Client 7. Research Design and Data Analysis Support 8. Bias, Representation, and Cultural Sensitivity 9. Crisis, Risk and Suicide: The Definitive Frontier of Artificial Intelligence 10. Therapeutic Chatbots and Digital Mental Health Tools 11. Supervision, Professional Development and Administrative Workflow 12. End-to-End Integration: Ethical AI Use Policy
Unit 4 / 12

Supporting Scale and Test Interpretation

Gains:

  • Ability to use artificial intelligence as cut-off score, subscale and linguistic explanation support in interpreting scale and test results
  • Ability to limit AI output by maintaining that diagnosis arises from clinical judgment and multiple data, not from scale score
  • Ability to consider test copyright, norm context and cultural validity risks in artificial intelligence interpretation

In psychology, scales and tests (such as anxiety scale, depression inventory, personality test, intelligence test) are tools that try to translate a person's inner world into numbers. But a number alone never describes a person. Someone who scores "high" on a depression scale may actually be in a temporary state of grief; Someone with a “low score” may be hiding their symptoms. That's why scale interpretation is an art: combining the number with clinical judgment, history, and context. Artificial intelligence (AI) can be an aid in this process—reminding what a subscale measures, explaining a score in plain language, refining a report sentence—but it will never provide the diagnosis and interpretation. In this unit you will learn how to use AI as a safe support in scale interpretation.

What AI can and cannot do in scale interpretation

AI can: explain the general structure and subdimensions of a scale in plain language; explain in general terms what a score "could mean"; improving the language of a report; drafting a simple text to explain to the client; Describing the overall pattern between different subscale scores.

What AI cannot do: derive a diagnosis from a score; deciding what the cutoff score (the threshold that marks the “clinical level” limit on a scale) means for that person; to ensure the cultural/linguistic validity of the result; Knowing the norms of the test (in which group and with what averages that test is standardized). AI can also generate misinformation by “pretending to remember” copyrighted test items.

Caution: The scale score is a piece of data, not a diagnosis. Diagnosis; It results from the combination of interview, history, observation, and multiple measurements with clinical judgment. Ask the AI ​​"what is the diagnosis based on this score?" To ask is to invite the agent to do a job that he cannot do.

Cut-off score and the trap of the norm

Each measure has a norm: that is, that test is standardized in a specific sample (for example, in a specific country, age and language group). The same cutoff score in another culture may be misleading. For example, a "high" threshold for an anxiety scale developed in one country may make normal emotional expression seem pathological in a different culture. AI does not know this context; It just does a mechanical mapping like "high score = high anxiety". It is the expert's job to weigh the interpretation within the framework of cultural validity.

Additionally, the cutoff score is probabilistic, not categorical. “1 point over the limit” is not the same as “20 points over the limit”; There is a margin of measurement error. AI tends to divide a score into strict categories of “mild/moderate/severe”; whereas true interpretation knows that these boundaries are blurred.

Step by step: AI-powered safe scale interpretation

  1. Calculate/verify raw score and subscales yourself. Don't leave the scoring to AI; AI can also make arithmetic errors.
  2. Ask the AI ​​for structure clarification. “What does this subscale generally measure?” ask; This is general information, not personal commentary.
  3. Add the context yourself. Incorporate stories, observations, and other data into your own judgment, not the AI.
  4. Use AI at the language level. Have the AI ​​draft your explanation to the client or report, then correct it for clinical accuracy.
  5. Question cultural validity. Is this scale appropriate for this client's culture/language? You ask this question that AI missed.
  6. Establish diagnosis from multiple data. The score is interpreted as part of the table, not in isolation.
Tip: Use the AI ​​like an editor who “translates my comment into understandable language” rather than “interprets the score”. Let the interpretation itself come from you; Let AI just make it beautiful.

three mini cases

Case 1 — The danger of the mechanical interpretation. An expert gives the result of a depression inventory (28 points) to the AI ​​and says "interpret". AI says "severe depression, treatment required." However, the client lost a relative two weeks ago; The score reflects a grief response. Without clinical context, the AI ​​label of “severe depression” is misleading and damaging. When the expert adds context, the interpretation changes completely.

Case 2 — The right support. A psychologist self-scores 5 subscales of a personality inventory, then tells the AI ​​"write in plain language what each subscale generally measures, prepare a draft text that I will explain to the client; DO NOT write a diagnosis." AI produces a clear outline; The psychologist corrects and uses clinical accuracy. Approximately 40 minutes of writing work is reduced to 10 minutes, and the comment comes from the expert again.

Case 3 — Cultural validity. A client scores high on a social anxiety scale developed in another country. AI suggests “marked social anxiety disorder.” The specialist realizes that the norms of the scale do not fit the cultural context of the client and that some items pathologize normal shyness in that culture. The score is the same, but the interpretation changes when passed through the filter of cultural validity.

Copiable prompts and templates

I want to explain the subscales of a psychometric scale to the client. For the subscale names below, explain in plain and non-stigmatizing language what each of them generally measures, WITHOUT WRITTEN A DIAGNOSIS or MAKING PERSONAL COMMENTS. Subscales: [list]

Translate my draft comment below into plain language the client will understand. CHANGE content, ADD new clinical claim; just make the language understandable and non-judgmental. Adding a diagnostic statement.Draft: [my own comment]

This scale [scale name/type] is used in a different cultural/language context. Write a checklist of general risks (norm context, item appropriateness, wording difference) that I should pay attention to in terms of cultural and linguistic validity.

In the report paragraph below, flag statements that are overly precise (“definitely,” “severely,” “evidence that”) or imply a single-score diagnosis, and suggest that I rewrite it in more measured, multi-data-aware language. Paragraph: [text]

Weak prompt / Strong prompt

Weak prompt: "Beck depression score is 28, what is the diagnosis?"

This prompt forces the AI ​​to make a diagnosis, even from a single score and without context; It is both wrong and unethical.

Strong prompt: "Explain in plain language which areas a depression inventory generally screens for. DIAGNOSIS. Also explain with a list of caveats why a high score alone does not mean a diagnosis (such as grief, temporary stress, measurement error, cultural context)."

This prompt puts the AI ​​in the educational/explanatory role; leaves the diagnosis and interpretation to the specialist.

Common mistakes

  • Making a diagnosis from the score. Translating a single number into a clinical verdict; whereas diagnosis arises from multiple data.
  • Mistaking the cut-off score as the definitive limit. Ignoring measurement error and context and saying "above the limit = patient".
  • Bypassing cultural validity. Applying a scale developed in a different culture without asking for context.
  • Leaving the scoring to AI. AI may make errors in arithmetic and item matching; scoring must be verified.
  • Printing copyrighted test items to AI. AI may “remember” test items incorrectly; Additionally, many tests are copyrighted.

In summary

The scale and test score are a piece of data, not a diagnosis. Explaining the AI ​​subscale structure helps the client prepare understandable language and improve the report sentence; but the diagnosis, cut-off score interpretation and cultural validity decision come from the expert through clinical judgment. Validate the scoring, add context yourself, read the cutoff score probabilistically and culturally sensitive. Use AI not as a commentator, but as an editor who polishes your comment.

Application task

Choose a sub-dimension of a scale you use. First, write a paragraph of your own clinical commentary. Then give the AI ​​this paragraph saying “translate to plain language, add clinical claim” and compare the output: At what points did the AI ​​intervene in the content, where did it just correct the language? Also ask the AI ​​about the risks of using the same scale in a different culture and come up with a checklist.

checklist

  • [ ] I verified the scoring myself, I did not leave it to the AI.
  • [ ] I established the diagnosis not from the score, but from multiple data and reasoning.
  • [ ] I read the cutoff score as probabilistic and context sensitive.
  • [ ] I questioned the cultural/linguistic validity.
  • [ ] I used AI as a language editor, not an interpreter.
  • [ ] I have moderated the overly precise statements in the report.
Previous Next