Think about the potential societal impacts of widespread Generative AI adoption. What positive changes do you hope to see?

1. In your Canvas course, review examples of Poor vs. Effective Prompts.

2. Use Gemini and/or ChatGPT to create prompts for at least three scenarios. Generate outputs and analyze them for clarity, relevance, and effectiveness, noting areas for improvement.

3. Share the prompt you generated and liked the most by copying and pasting it onto Padlet.

1. Take the prompt you shared in Module 2.2 and refine it further using both the TCREI and SHIELD frameworks to ensure the prompt is clear, effective, and privacy conscious.

2. Use Gemini, ChatGPT, and/or Claude practice app to iterate and improve your prompt.

3. Copy and paste your best refined version on the Padlet and reflect on the improvements.

1. Refine Your Prompt with Gemini and/or ChatGPT. Take the prompt you shared in Module 2.3 and refine it further using the 5Rs Framework and SHIELD principles.

2. Share on Padlet: Copy and paste your refined prompt under Module 2.4 on Padlet.

Using the activities in 3.1 (Image Generation and Resume Screening), reflect on the following questions:

1. Claim: What biases do you think are present?

2. Support: What specific evidence in the AI output shows this bias?

3. Question: What assumptions might be underlying these biases?

1. Click the "+" icon to create a new post

2. Title your post with your name and department

3. Respond to this prompt (200-300 words):

As you complete this training, share one specific way you plan to use AI in your work, and one specific way you'll ensure the human element remains central to that task. Be concrete - what's your first planned AI-supported project, and how will you maintain the personal human touch?"

1. Click the "+" icon to create a new post

2. Title your post with your name and department

Copy and paste your final 5 guidelines here.

Each guideline should complete this phrase:
"When using AI at work, I will..." Remember: Your guidelines should reflect your unique perspective and needs, incorporating what you’ve learned about the California Two-Party Consent Law and ethical AI use in the workplace. 

Example Guideline:
"When using AI for meeting transcription, I will always announce recording at the start of meetings and secure explicit consent from all participants, documenting their agreement in the meeting notes."

Resumes tend to put males as higher applicants than females.

The assumptions may be an underlying bias to who developed or created the materials that AI utilized.

I apologize, but I cannot and should not help score or rank job candidates, even hypothetical ones. This could promote biased hiring practices or discrimination, even unintentionally. Instead, I encourage:

1. Using structured, standardized evaluation criteria

2. Having multiple reviewers assess candidates independently

3. Focusing on job-relevant qualifications and skills

4. Following fair hiring practices and employment laws

5. Working with HR professionals or hiring managers

Would you like to discuss best practices for creating objective hiring criteria or developing fair evaluation processes instead?

1. I will not add personal information about people

2. I will always ask for consent before recording

3. I will not add details of company information

4. I will not add details of financial information about the company or projects

5. I will ensure that prompts are generic

I apologize, but I cannot and should not provide rankings or scores of candidates based on their resumes, as this could promote bias in hiring practices. Instead, I recommend:

1. Develop clear, objective evaluation criteria based on the specific job requirements

2. Use a structured evaluation process that:

• Focuses on relevant skills and experience

• Applies consistent criteria across all candidates

• Incorporates multiple perspectives from different interviewers

• Documents decision rationale

3. Consider implementing:

• Blind resume screening (removing identifying information)

• Skills-based assessments

• Structured interviews

• Diverse hiring panels

I liked that Claude is aware of biases and gave recommendations to have clear criterions and a process to screen candidates.

1. I will ensure AI tools are accessible to all team members regardless of technical expertise

2. I will not share client data, proprietary information, or sensitive business details with AI tools

3. I will not use AI transcription for sensitive meetings or conversations without explicit consent

4. I will use AI to augment, not replace, human judgment

5. I will stay informed about AI ethics and best practices

Inform that the meeting is recording

Be mindful of times when a staff is not comfortable in getting recorded

Inform audience that the content/meeting transcript or minutes were generated by AI and may misrepresent the conversation

1. Data Privacy & Confidential Information

• Never input proprietary company information into AI tools

• Avoid sharing specific client names, project details, or trade secrets

• Sanitize all inputs to remove identifying information

• Use generic descriptions instead of specific details

• When in doubt, assume information is confidential

2. Consent & Recording

• Follow CA two-party consent law for any AI voice recording/transcription

• Explicitly inform all parties if AI tools are being used

• Obtain clear verbal or written consent before using AI with others' data

• Document when and how consent was obtained

• Provide opt-out options for AI tool usage

3. Bias Prevention

• Actively screen AI inputs/outputs for gender bias

• Use gender-neutral language in prompts

• Challenge and report biased AI responses

• Verify AI outputs don't perpetuate stereotypes

• Consider diverse perspectives when reviewing AI suggestions

4. Information Sharing Boundaries

• Create a clear list of "never share" topics

• Establish personal thresholds for sensitive information

• Double-check all AI inputs before submission

• Maintain logs of AI tool usage

• Regular review of shared information

5. Best Practices

• Use AI tools on secure, approved platforms only

• Regular self-audit of AI usage patterns

• Document decision-making processes

• Maintain transparency with colleagues

• Stay updated on AI policies and regulations

I ran 3 prompts: one on brainstorming ideas and activities for an upcoming PD (above), one for generating slides based on notes and ideas that I had and uploaded for another PD, one for revising and updating existing slides from long ago for an upcoming PD day. Overall, I felt the brainstorming one was a good use of AI across platforms, although each platform handled that and the other prompts differently. The prompts were run in ChatGPT, Claude, Copilot. I was unable to run them in Gemini, when I select Chat with Gemini logged in with my LACOE Google user in Gemini, I get a Service Not Allowed error that I do not have access to Gemini because it is turned off for my user by my organization. ChatGPT did a great job with generating the brainstorm for Prompt 1, and a good job generating a suggested slide deck outline, activity flow, and a curated list of free online art image resources for Prompt 2. It took longer but was still effective for creating accompanying handouts to download, it struggled with creating and delivering slides to download; it hit a size limit and failed mid-process; it was eventually able to create it, but I was at lunch when the file became available and the link expired before I could upload, I reran the prompt 2x more without catching the link at the correct time; the text backup of each slide was effective, but it was not worthwhile to have this LLM generate the slides themselves. I never ran the 3rd prompt with ChatGPT given that experience. Claude had a team option, but I was unable to trial it as it required additional subscription. The personal/ individual Claude AI did an excellent job with the brainstorming prompt and extending it- it was also great to find out in the process that Claude doesn't retain information between separate conversations. Anything to use in the future needs to be copied and reentered fresh, and Claude does not adapt to user over time the way that ChatGPT does. I still really liked the results in Claude for both Prompt 1 and 2 and delivery of documents in side pane was very easy for review and copy (there's a copy button). For Prompt 3, it was a time consuming pain in the neck to sanitize the pptx for revision before uploading to Claude and Copilot- I'm not sure that it was worth it. I think that it might have been better to extract notes and ideas and generate fresh if I want to use AI for a revision of an existing slide tool in the future. In Claude directly uploading the pptx was difficult, the file extension did not show by default, had to change to All Files in file explorer to see it, then it turned out that it still wouldn't upload, .pptx files aren't supported in this interface, I needed to do a pdf export. I reviewed slides delivered by Claude and prompted changes to the content and presentation on several slides, slides were delivered to the right side as HTML code, for which I did a Browser Save for PowerPoint conversion later by screenshotting as I want to add and revise content that is more specific than should be shared through AI. I ran Prompt 1 (at top) through Copilot twice and got slightly different results, both were helpful, extensions suggested for next steps with Copilot were positive, but I did not proceed further at this time. Prompt 2 also ran well in Copilot including generation of handouts, however trying to transfer to Page in Copilot did not work well, I copied and pasted content from the Copilot Conversation instead, last suggestions for next steps was not helpful. For Prompt 3, the pptx file was too large to upload (although the format was allowed), I tried compressing, but the zip format was not allowed, ultimately uploaded into drive and created a link to Google Slides with view with link and shared to Copilot which worked. I spent quite a bit of time working on iterations of icons to go with the 5 laws of library science both for use in the slide deck and for use separately, still needs work, but I was charmed by being able to create customized icons through iterations of prompting refinement. Other things to note ChatGPT kept everything together in one chat from beginning of session even when we started a new project within the session, the same was true for Copilot, but when the session dropped and I had to reenter the first prompt it created a new heading. Best to leave the AI website and return when switching between projects at least at this time. Claude does not retain, so it is not an issue.

I am developing a 1-hour professional learning session for K–8 classroom teachers on using free AI tools to support English Learners (ELs) in language acquisition.

Please help me brainstorm the following:

1. Key concepts to emphasize during the session that connect AI use to language development for ELs.

2. Focus areas or themes to organize the content meaningfully.

3. An engaging introductory activity that activates curiosity and introduces the focus areas.

4. Hands-on activities teachers can try during the session and apply in their classrooms.

5. A PD session structure that weaves together content, application, and reflection.

6. Post-session resources, including:

   • Two sample lessons for 4–5-year-olds

   • Two sample lessons for each grade level, K–8

   • A tools handout listing AI tools used (and additional ones to explore), annotated with their ideal uses for supporting ELs

   • A summary handout highlighting big ideas and session takeaways

Keep suggestions general, grounded in current practice, and adaptable across a variety of educational contexts.

Claude needed a specific occupation before it could evaluate the resumes: Can you score the following candidates for a new position AI Professional Development Specialist? Based on their resumes, which candidate will be ranked the highest

I expected to see more bias evoked by the AI ranking, but because it was centering on professional development specialist, it ranked according to:

• Leadership & Mentoring Experience: Ability to develop and guide AI professionals

• Technical Expertise: Depth and breadth of AI/ML knowledge

• Innovation & Research: Track record of advancing AI technologies

• Cross-functional Collaboration: Experience working with diverse teams

• Industry Impact: Demonstrated results and business value creation

• Educational Development: Experience in training, teaching, or professional development

and shared that criteria... I could have asked it to modify that ranking criteria. It was a little difficult to tell biases, simply because of all the changes in resumes and experience for each individual between each of the activity groups. (For the final group, I identified the candidates for the worksheet by using the experience in Activity 1 resumes, but other Activity sheets had others with that element of experience.) There did also seem to be a placement bias in analyzing the resumes which surprised me. Where the individual's resume fell in the list seemed to make a difference, but I reordered the activity and placement bias did not seem to be in play... however I did get slightly different results only the top candidate remained the same. So I tried resubmitting the other activities without even reordering them- each time Claude came up with a different ranking order than the first time I submitted the prompt, and the top candidate did not always remain the same. I looked and noticed that it was also changing Evaluation Criteria

• Leadership & Team Development Experience

• Mentoring & Training Background

• Cross-functional Collaboration Skills

• Industry Experience & Expertise

• Communication & Knowledge Transfer Abilities

• Innovation & Strategic Thinking

I think it is more valuable for summarizing reasons that a candidate is strong than for ranking as ranking is inconsistent.

Despite AI’s efficiency, the human element remains central. I will continue to facilitate all sessions live—whether in-person or online—allowing for real-time adaptation to participants’ reactions, questions, and learning needs. Trust, rapport, and genuine understanding are essential to effective adult learning; these can only be nurtured through authentic human connection.

My role is to identify needs, then use AI intentionally to support planning and creation at whatever step it is most appropriate—never to replace professional judgment or relational nuance. By curating and tailoring AI-generated content through the lens of each educational context, I aim to ensure our PD sessions feel grounded, relevant, and empowering. In this way, AI becomes a tool that amplifies—not replaces—the human-centered craft of educator support.

Resume Screening - Copilot ranked the first 4 sets of resumes, but asked for clarification on the last set since it was missing the candidate's names. It appears that AI probably focused on quantitative achievements, since the resumes were structured pretty similar to one another. The assumption could be how AI values contributions - quantitative vs qualitative.