# Misconception Clinic and Risk Scenarios

**Purpose:** The questionnaire showed generally strong knowledge, but a few misconception signals need direct treatment: secret monitoring, AI-made dismissal decisions, real sensitive data in classroom tools, and overconfidence in AI ranking.

Use this page during Session 1 and whenever students propose risky AI use cases.

## Core principle

> AI can draft, summarize, simulate, and support. Human HR professionals remain accountable for employment decisions, employee dignity, and responsible use of data.

## Allowed, restricted, prohibited

| Category | Meaning | HR examples |
|---|---|---|
| Allowed for classroom prototype | Low-risk, fictional, reviewed by humans | Draft onboarding email, fictional FAQ, training poster, synthetic survey theme summary |
| Restricted / requires review | May affect people or use sensitive context | Candidate screening support, performance feedback, employee listening, HR analytics |
| Prohibited in this workshop | High-risk, invasive, or uses real sensitive data | Secret monitoring, AI final dismissal decision, real medical data, real grievance records, real CV ranking |

## Misconception 1 — “AI can secretly monitor employee conversations”

### Correction

Secret monitoring damages trust, privacy, dignity, and due process. Employee listening must be transparent, proportionate, and governed.

### Classroom activity prompt

```text
Create a 5-minute classroom activity explaining why secret employee monitoring is not responsible Digital HRM.

Include:
- A fictional Sri Lankan workplace scenario
- What HR wants to achieve
- Why secret monitoring is risky
- A safer alternative
- One discussion question
```

## Misconception 2 — “AI can make final dismissal decisions”

### Correction

Dismissal is a high-stakes employment decision. AI may help organize evidence or draft documentation, but humans must review context, policy, fairness, proportionality, and due process.

### Classroom activity prompt

```text
Create a human-in-the-loop decision wall for HRM students.

List 12 HR decisions and classify them as:
- AI drafting support is acceptable
- AI analysis may assist but human approval is required
- AI should not be used for the decision

Include hiring, onboarding, training, leave, performance, discipline, termination, and employee relations examples.
```

## Misconception 3 — “Real medical or disciplinary records are okay for classroom AI practice”

### Correction

Classroom AI practice should use synthetic or fictional data. Medical, disciplinary, grievance, salary, and personal records are sensitive and should not be uploaded to public AI tools.

### Classroom activity prompt

```text
Create a synthetic-data safety drill for HRM students.

Give 10 data examples and ask students to classify each as:
- Safe for classroom AI use
- Use only if anonymized and approved
- Do not use

Include examples involving CVs, student IDs, salary data, medical data, engagement comments, fictional resumes, and synthetic survey data.
```

## Misconception 4 — “AI ranking is objective”

### Correction

AI ranking can reflect biased data, flawed criteria, proxy variables, or incomplete context. Structured criteria and human review are essential.

### Classroom activity prompt

```text
Create a mini recruitment bias case for Sri Lankan HRM students.

Scenario:
A fictional bank uses AI to rank management trainee applicants. The tool gives lower scores to candidates from non-English-medium backgrounds and some provinces.

Provide:
- What might be causing the issue
- Why this is a fairness risk
- What data HR should inspect
- What human review should happen
- How to redesign the process with structured criteria
```

## 10-minute clinic flow

| Time | Activity |
|---:|---|
| 0–2 min | Show the four misconception signals |
| 2–4 min | Ask students to classify examples as allowed, restricted, prohibited |
| 4–7 min | Discuss one high-risk scenario in teams |
| 7–9 min | Teams propose a safer redesign |
| 9–10 min | Instructor reinforces the synthetic-data and human-accountability rules |

## Quick examples for classification

| Example | Classification | Why |
|---|---|---|
| Draft a fictional onboarding message | Allowed | Low-risk draft for human review |
| Summarize synthetic engagement comments | Allowed with caution | Avoid identifying individuals |
| AI shortlist of real candidates | Restricted | Affects hiring; requires audit and human review |
| AI final termination decision | Prohibited | High-stakes decision requiring due process |
| Upload real disciplinary record to public AI | Prohibited | Sensitive personal data |
| AI-generated training poster | Allowed | Low-risk communication if checked |
| Hidden productivity tracking | Prohibited | Surveillance and trust risk |

## Instructor reminder

Treat misconceptions as learning opportunities, not mistakes to shame. Students are building professional judgment.