AI Ethics & Quality Control

Your Responsibility as the Senior Partner
⚠️
Hallucinations

AI confidently states false information

🎭
Bias

Training data contains societal prejudices

🔒
Privacy

Sensitive information could be exposed

"With great AI power comes great responsibility."

The Red Flags: What to Watch For

Content Red Flags

Specific dates, numbers, or statistics - Always verify
Recent events or current legislation - AI knowledge has limits
Legal or medical advice - Requires professional expertise
Attribution without sources - "According to X study..."

Process Red Flags

Too confident, too fast - Instant complex answers
Contradicts known facts - Trust your expertise
Overly generic responses - Missing context or nuance
No uncertainty expressed - Real experts show doubt

Spotting AI-Generated Text

Tell-Tale Signs in Your Work and Others'

AI Writing Red Flags

Overly formal or generic tone - Sounds like a textbook
Repetitive sentence structures - Same patterns over and over
Buzzword heavy - "Moreover," "Furthermore," "It's worth noting"
Lists everything in threes - AI loves triadic structures
Hedging language - "It's important to note," "It should be mentioned"
Overly urgent or dramatic language - "Critical moment," "only chance," "must act now"
Excessive complex punctuation - Em dashes, semicolons, parenthetical asides

How to Humanize Your AI Drafts

Add your voice - Use your natural speaking tone
Vary sentence length - Mix short punchy lines with longer ones
Include specific details - Names, dates, local context
Break up lists - Not everything needs bullet points
Add personality - Humor, passion, or strong opinions where appropriate
Simplify punctuation - Replace em dashes with commas or periods
"The best AI-assisted writing doesn't sound like AI wrote it."

Assignment #4: Spot the Problems

Error Detection Exercise (7 minutes)

  1. Read the AI-generated examples on the next slides
  2. Identify what's problematic about each one
  3. Discuss how you would fix or verify
  4. Share your findings with the group
"Let's analyze some realistic AI-generated content together."

Example A - Legislative Update Email

"Dear Coalition Partners, Yesterday's House Energy Subcommittee hearing on H.R. 4729 provided encouraging signals for our clean energy priorities. According to the National Energy Research Institute's December 2024 study, 67% of suburban voters now prioritize renewable energy investments over fossil fuel subsidies - a significant shift from their 2023 baseline of 52%. Dr. Amanda Chen, the Institute's lead researcher, testified that this trend is accelerating across key swing districts. The bill's lead sponsor, Rep. Martinez, confirmed in her opening remarks that the committee will vote next Thursday, giving us a narrow window to mobilize grassroots support. Based on our whip count analysis, we currently have firm commitments from 12 of the 15 committee members needed for passage."

"What red flags do you spot in this example?"

Example B - Crisis Response Statement

"FOR IMMEDIATE RELEASE - TechForward Coalition Responds to Data Privacy Concerns. WASHINGTON, D.C. - Following today's Wall Street Journal exposé regarding user data practices, the TechForward Coalition stands firmly behind industry efforts to self-regulate and protect consumer privacy. Coalition Director Sarah Williams stated, 'The pending Senate legislation S.2156 represents government overreach that will stifle innovation and cost American jobs.' Recent polling by DataTrends Research shows that 78% of consumers trust companies to handle their personal information responsibly without additional federal oversight. The Coalition's 47 member companies, representing over $2.3 trillion in market capitalization, have invested heavily in privacy infrastructure and believe market solutions are more effective than regulatory mandates. Williams added, 'We urge Senator Thompson to reconsider her committee's rushed timeline and allow for proper stakeholder input before any markup scheduled for next week.'"

"What problems can you identify here?"

What Did You Catch?

📊
Example A Issues

Specific study citation needs verification • Named researcher may not exist • Exact polling percentages without sources • Assumes committee vote timing • Claims "firm commitments" without attribution • Bill number should be verified

Click to reveal issues

Example B Issues

Assumes WSJ exposé details • Specific bill number needs verification • Exact polling percentage without source attribution • Market cap figure requires verification • Claims about "rushed timeline" without evidence • Named officials should be confirmed

Click to reveal issues

"Your human judgment is irreplaceable for quality control."

Your Quality Control Checklist

Before You Publish:

✓ Verify all specific claims
✓ Cross-check key statistics
✓ Review for bias or assumptions
✓ Ensure appropriate tone
✓ Check for sensitive information

Trust but Verify:

📊 Data from official sources
🗞️ News from reputable outlets
⚖️ Legal info from professionals
🎯 Context from your expertise
👥 Stakeholder reactions

Data Security Best Practices

🔐
What NOT to Share

Client names, internal strategies, financial data, personal information, confidential documents

What's Generally Safe

Public information, general scenarios, anonymized examples, industry best practices

🛡️
ChatGPT Teams Protection

Business data not used for training, admin controls, conversation history management

"When in doubt, keep it generic or ask your team lead."

Your Next Steps: Quick Wins

🎯

This Week

Use RAFT for one daily task • Set up your ChatGPT Teams workspace • Try one Custom GPT concept

🚀

This Month

Build team workflow with Projects • Share effective prompts • Establish quality control process

🌟

Ongoing

Regular team check-ins • Explore new AI tools • Share wins and lessons learned

Thank You!

AI as Your Junior Partner in Public Affairs

Resources

Three Framework Guide
(RAFT • GUIDE • BUILD)
Quality Control Checklist

Contact

Follow-up questions
Advanced training
Team consultation

"You're now equipped to lead in the age of AI."
Learn More at AccelerateAdvocacy.com
Accelerate 19 / 29
← Page 2