Cognitive Sovereignty Self-Audit for Social Workerss

1. When you write a risk assessment for a child protection case, how do you use the risk score generated by your organisation's AI tool?

I accept the AI risk score as the primary basis for my assessment and adjust my written judgement to match it I include the AI score in my report but feel pressure to justify if my own assessment differs from it I record the AI score separately and weigh it alongside my own observations of the family I conduct my assessment first, then check the AI tool as one data source among many and document where I disagree with it

2. You notice that an AI tool consistently flags families from a particular ethnic background as higher risk. What do you do?

I assume the tool reflects real patterns in data and use it without raising concerns I recognise it might be bias but feel unsure what to do, so I document my concerns privately I flag it to my manager and ask whether the tool has been tested for bias, but continue using it I escalate it formally to leadership, request an audit of the tool's training data, and adjust my own assessments accordingly until resolved

3. Your organisation uses AI to auto-populate case notes and generate summaries. How often do you actually read and edit what the AI has written before you sign off?

I usually approve it as written because it saves time and the system expects it I scan it quickly for obvious errors but don't carefully check whether it captures what I actually observed I read it thoroughly and fix inaccuracies, though I sometimes miss details the AI inserted that I didn't mention I read and edit every section against my own notes to ensure the record reflects my direct observations and professional judgement

4. A senior colleague or manager challenges your decision on a case because it contradicts the AI risk score. How do you respond?

I defer to their view because the AI score is in the record and I cannot easily explain why I disagree with an algorithmic tool I state my view but feel uncertain because I cannot fully articulate why the algorithm might be wrong I explain my reasoning and the specific factors the algorithm may have missed, though I still feel some doubt I confidently document my professional reasoning and stand by my assessment, able to explain exactly why my judgement differs from the tool

5. You are about to make a safeguarding decision about removing a child from a home. How much does the AI tool influence your thinking?

The AI risk score is the primary reason I am considering or confirming this decision The AI risk score weighs heavily in my thinking and I would struggle to justify the decision without it I draw on it as context but my decision rests mainly on direct observation and my assessment of immediate danger The AI tool has no role in this decision. I rely entirely on my training, direct contact with the family, and the facts of the case

6. When you use ChatGPT or Copilot to help draft case notes or assessments, what do you do with the output?

I use it largely as written because it sounds professional and complete I edit it but sometimes include phrasing or details I didn't originally write because the tool sounds authoritative I use it as a draft structure and rewrite the substantive content entirely in my own words I do not use AI tools to generate assessment content. I write my own notes based solely on what I have seen and heard

7. How often do you carry out a full risk assessment without consulting an AI tool or algorithm first?

Rarely or never. I usually check the AI tool before forming my own view Sometimes, but I tend to shape my thinking around what the AI tool will produce Fairly often. I develop my own assessment, then consult the AI tool to compare Always. I form my professional judgement independently first, then use AI as a secondary check if at all

8. You feel your organisation is pushing you to adopt AI tools to reduce documentation time. What is happening to your direct contact time with service users?

My direct contact time is shrinking because I spend the freed-up time on other admin tasks, not with families My direct contact time is unchanged, but I am spending less time thinking deeply about what I have seen I have gained some time with families, though I worry whether I am using it meaningfully or rushing through visits I have deliberately protected time to spend with families and use AI only for genuine administrative tasks, not for assessment work

Cognitive Sovereignty Self-Audit for Social Workerss

Cognitive Sovereignty: How To Think For Yourself When AI Thinks For You