Every time it’s evaluation season (now), I think about how anxious I get knowing this task is big and lurking. If my feedback to teachers is valuable, they can use it to develop their skills, bringing better teaching and learning to their students. Yet, the day-to-day humdrum of a busy school building, parental needs, and budget constraints increase my work load, and more than often, get in the way.
Balancing this task with being a visible leader is a delicate challenge. How can I save time, give better qualitative feedback, and help my school community through me being accessible?
Enter AI…
Let’s face it, teacher evaluations are one of the most critical, time-consuming, administrative tasks school leaders face, yet also warrant careful documentation, objective analysis, and the crafting of nuanced, actionable feedback. The emergence of large language models such as Gemini, ChatGPT, Meta, and others, offers an invaluable opportunity to streamline the narrative drafting and synthesis components of the evaluation cycle, improving quality and reducing time on the task.
Don’t let me sell you too hard–AI cannot replace the evaluator. The true power of AI lies in an integrated approach in which human judgment and professional discretion remain the cornerstones. Recognizing this helps me provide you with the roadmap to save countless hours while increasing the quality of your feedback.
This guide outlines a three-phase process for integrating AI into your evaluation process, ensuring integrity, objectivity, and a significant boost in efficiency. It’s not an end-all, and you can take from it what you want, adapt the other parts and add your own ideas. It provides you with rational, ethical reasoning, and I’m certain, a few prompts you will use again and again in your own school leadership routine.
Phase 1: The Analog Foundation (Evidence Collection & Preparation)
The quality of the AI-generated draft hinges entirely on the quality and objectivity you put into it. This phase is deliberately human-centric and serves as the foundation of the entire process.
1. Collect Objective Observation Notes: The process begins with traditional, human-led observation. I write all my notes in a factual, objective, evidence-based manner. I’m sure you do this part already. These should remain “dry, straight facts.” At this stage, I avoid any subjective interpretation or judgment.
2. Standardize and Input Evidence: Transfer the raw, factual evidence into your official evaluation system. An important step here is a thorough check for grammar and structural clarity. Be sure that the foundational data is properly edited before engaging the AI.
3. Prepare Evidence for AI: Engage in the process of removing or obscuring personally identifiable information, which is the most important security feature before any data is entered into an external AI tool. It may seem like a waste of time, and now you may want to give up on the AI, but don’t. I still save at least half as much time on an output on the other side of the process by taking this step. Remove the teacher’s name, names of collaborative partners, student names, and specific dates/times that could uniquely identify the educator. The evidence is now “blind” and ready to be inputted into the AI chatbot.
Phase 2: The AI Drafting & Narrative Enhancement
With a body of clear, objective text evidence, the AI chatbot is now ready for its part: drafting cohesive, well-phrased narratives. Using the chat history to maintain context is key to building a comprehensive final summary.
1. Generate Feedback on Pre-Observation Discussion: Use a specific, focused prompt to generate initial feedback.
- Example Prompt I Use: Use this teacher discussion of her lesson plan for an evaluation by the principal, to write 3 sentences of feedback, from the principal about the lesson:
2. Draft Narrative from Post-Observation Reflection: The AI excels at taking the teacher’s self-reflection and re-framing it into the third-person narrative required for the principal’s perspective.
- Example Prompt I Use: Use this teacher reflection of the lesson to write in the third person, from the principal’s perspective in 3 sentences:
3. Capture Ongoing Evidence History: Context matters. Continue to input all relevant, anonymized evidence throughout the evaluation cycle using the same chat session. This ongoing history provides the comprehensive context the AI needs for the final summary step. By the way, you can always go back to a chat session in your AI chatbot history (here is how to do that in one LLM).
4. Craft Areas for Growth: A significant benefit of using AI is its ability to draft feedback in a consistent, objective framework.
- Example Prompt I Use: For areas to develop, use the entire history of evidence to write 2 sentences for areas of growth, framed in the positive (as in continue to…)
5. Synthesize the Performance Overview: This is where the time savings are most obvious. By utilizing the entire chat history—all the inputted evidence—the AI can generate a synthesized summary paragraph that links disparate pieces of evidence into a coherent narrative.
- Example Prompt I Like to Use: Use all of the above inputted evidence to write a 5-sentence overview of the teacher’s performance.
Phase 3: The Human Refinement (Professional Discretion & Quality Control)
Remember, the output from the AI is a draft—not the final document. The ultimate responsibility and authority rests with you. This phase ensures the evaluation is accurate, fair, and, most importantly, personal, or what I like to call personification.
1. Review, Revise, and Personalize (The Final Check): This is the stage when your professional discretion comes through. The evaluator must revise all components of the AI-generated text, particularly when the AI drew conclusions or made interpretive statements.
2. Add Personal Context: Integrate that personified information that only you uniquely know about the teacher (e.g., their long-term goals, specific mentoring relationships, or particular contributions to the school culture) to make the feedback more authentic and meaningful.
3. Quality Control: Scrutinize the AI’s output rigorously for:
- Strange Wording/Tone: AI can sometimes be overly academic or use passive voice. Use your voice throughout, touching up this style.
- Redundancies: A common AI tendency is to repeat points or phrases, so look for and remove these.
- Content Conflict: Ensure the output does not counter your professional judgment or your school’s established evaluation criteria. (For example, I use the Danielson Framework).
- Alignment: Directly align the drafted narrative to the specific domain and component language of your evaluation system. Here, you can even ask the AI where in your evaluation framework these components are best suited, and of course use your professional judgement both with AI support and on your own.
Using the integrated, three-phase protocol, I have moved away from the tedious, time-consuming process of narrative generation. Instead, I get to focus my limited time on being objective and providing meaningful, personalized feedback, in less than half the time.
Like all new processes, this approach took me some practice so don’t give up after the first observation or two. You will see that as you get familiar with the process your time, and the quality feedback you give both get better!
Enjoy our content? Make sure to add Tech & Learning as a preferred source on Google to keep up with our latest news, how-tos, profiles, events, and more.