A faculty member uses AI to generate quiz questions and discovers later that half of them test trivial details instead of critical thinking. An advisor relies on an AI system to identify at-risk students and notices it consistently flags first-generation students and working parents, even when they're doing fine. A registrar uses an AI chatbot to answer student questions and learns it's been giving incorrect information about financial aid deadlines for three weeks.

Podcast Version created with NotebookLM

Technical knowledge tells you what AI can do. Ethics tells you what AI should do, and more importantly, what it shouldn't.

You can understand how machine learning works and still make decisions that harm students. You can know the difference between narrow AI and generative AI and still adopt tools that perpetuate bias. Understanding the technology is necessary but insufficient.

This module focuses on the ethical dimensions that matter most for CTCs: bias and fairness, privacy and data security, transparency and trust, environmental impact, and institutional responsibility.

These aren't abstract philosophy questions. They're practical considerations that affect real students, real employees, and real educational outcomes every day.

Click the expandable headers below to view the content for each section.

The AEIOU Ethos Framework: A Foundation for Responsible AI

The AEIOU Ethos Framework (Gutierrez, 2025) provides five essential principles that translate complex ethical concepts into practical, implementable strategies for community and technical college environments.

A - Accessible

AI systems must be usable by every student, faculty, and staff member—regardless of device age, internet speed, disability, or financial situation. That means designing tools that work on older hardware, support low-bandwidth use, and offer multiple interface options for users with disabilities.

E - Equitable

AI systems must actively reduce rather than replicate structural biases. That means regularly reviewing how the system performs for different populations, identifying gaps in outcomes, and building in safeguards or alternate paths to ensure fair experiences for all.

I - Inclusive

AI design and deployment must reflect the full diversity of students, faculty, and staff—especially from underrepresented groups. That means involving these voices in design, testing, governance, and correction of the system.

O - Open

AI systems should not operate as opaque "black boxes." Students, faculty, and staff should be able to see how decisions are made, ask questions, and raise concerns—and institutions must publish clear policy on the system's use, including how to appeal or seek review.

U - Universal

AI systems must work across diverse educational contexts, from rural to urban campuses and from technical programs to transfer pathways. This means designing flexible, interoperable tools that function effectively regardless of institutional resources or scale.

Try This Prompt:

"I'm considering adopting [specific AI tool] for my work at a CTC. Help me evaluate whether it's truly accessible, equitable, inclusive, open and universal. What specific questions should I ask about: device requirements, internet bandwidth, cost to students, compatibility with assistive technology, and availability during non-business hours?"

Bias, Fairness, and Algorithmic Discrimination

AI bias isn't a theoretical concern. Research published in 2024 by Gándara et al. found that student success prediction algorithms produce false negatives for 19% of Black students and 21% of Hispanic students. Translation: the AI incorrectly predicted these students would fail when they actually succeeded.

Think about the consequences. An advisor sees an AI prediction that a student will probably drop out. The advisor, trusting the system, doesn't reach out with support. The student who would have succeeded with that support never gets it. The prediction becomes self-fulfilling.

This isn't hypothetical. This is happening.

How Bias Gets Into AI Systems

Training Data Bias: If the data used to train an AI system doesn't represent the full diversity of students, the system performs poorly for underrepresented groups.

Example: An AI advising system trained primarily on data from residential students at four-year universities will give bad advice to CTC students juggling work, family, and school.

Historical Bias: AI systems learn from historical data that may reflect past discrimination. If previous admissions decisions or grading patterns contained bias, AI trained on that data perpetuates those inequities.

Example: If historically, students from certain high schools received less financial aid, an AI system might learn to recommend less aid for students from those schools, reinforcing the original disparity.

Design Bias: Assumptions built into AI systems by developers can introduce bias, especially when developers don't share the lived experiences of the people who will use the system.

Example: An AI scheduling system might assume "optimal" class times are weekday mornings, disadvantaging students who work those hours. The developers probably never worked full-time while attending college.

Try This Prompt:

"I'm using [specific AI tool] in my work. How can I check whether it might have bias that could disadvantage certain students? What warning signs should I watch for? Give me 3-5 specific things to monitor."

Real Examples from Higher Education

Automated Essay Scoring: Multiple studies show these systems exhibit bias based on writing style patterns associated with different cultural and linguistic backgrounds. Students who write in styles common in their communities may receive lower scores even when demonstrating strong analytical thinking.

Research on this has been published by Les Perelman at MIT and by the Conference on College Composition and Communication, among others.

AI Detection Tools: Studies show these tools have high false positive rates for non-native English writers and neurodivergent students. The tools flag as "AI-generated" what is actually human writing that doesn't match the statistical patterns the AI expects.

In 2023, multiple students at universities across the country were falsely accused of using AI based on detector results. Some faced honor code violations. Some appealed successfully. Some didn't.

Predictive Analytics: Early warning systems for identifying at-risk students have shown bias related to socioeconomic status and race. These systems often don't account for the support services available to students or changes in students' circumstances, leading to inaccurate predictions that can influence how advisors interact with students.

Try This Prompt:

"A student tells me an AI detector flagged their original work as AI-generated. They're multilingual and English is their second language. Help me understand: why might this happen? What should I do? How should I handle this situation fairly?"

Addressing Bias: What Actually Works

Diverse Data Collection: Training data must represent all students who will interact with the system. This means deliberately including data from working adults, multilingual students, students with disabilities, and part-time students, not just traditional full-time enrollees.

Bias Testing and Auditing: Before deploying an AI system, test its performance across different demographic groups. After deployment, continue monitoring for disparate outcomes. If the system performs differently for different groups, investigate why.

Diverse Development Teams: Include people with varied backgrounds in AI tool selection, testing, and oversight. They'll catch assumptions and problems others miss.

Human Oversight: Maintain meaningful human involvement in decisions that affect student outcomes. AI recommendations should inform decisions, not make them automatically.

The key principle: AI systems should reduce barriers for students, not create new ones.

Privacy, Surveillance, and Data Ethics

AI systems require data to function. In education, that means student data, employee data, and institutional data. Much of it is sensitive. All of it requires careful protection.

What Data Are We Talking About?

Modern educational AI systems can analyze:

Academic performance over time (grades, assignment submissions, quiz scores)
Learning behavior patterns (time spent on tasks, number of attempts, help-seeking behavior)
Writing style evolution (vocabulary, sentence structure, editing patterns)
Communication patterns (email response times, discussion participation, tone)
Predicted outcomes (likelihood of passing, dropping out, transferring, graduating)

Some of this data lives in systems you control (your LMS, your student information system). Some lives with third-party vendors (Turnitin, Proctorio, Zoom, various AI tutoring platforms). The distinction matters for privacy and security.

Try This Prompt:

"I want to use [specific AI tool] but I'm not sure what data it collects or who can access it. What specific questions should I ask the vendor or IT department before I start using this tool with students? Give me a list of essential privacy questions."

Key Privacy Concerns

Consent and Power Dynamics: Students can't really say no to institutional requirements. If your college requires using Turnitin or taking proctored exams through an AI monitoring system, students have limited choice. This creates an ethical obligation to be especially careful about which tools you adopt and how much data they collect.

Data Security: Centralized student data creates targets for malicious actors. AI systems that process data through cloud services may increase exposure. Every additional vendor handling student data is another potential vulnerability.

A 2024 breach at a major educational technology company exposed personal information for millions of students. Your college's vendor contracts should specify security requirements, but not all do.

Third-Party Access: Many AI tools require sending data to external companies. You need to know:

What data gets sent?
How is it stored and secured?
Who else can access it?
How long is it retained?
Can it be used to train future AI models?

Students (and faculty and staff) often don't know their data is being shared. They assume when they upload an essay to get AI feedback, it stays private. Often it doesn't.

Secondary Uses: Data collected for one purpose (advising) might be used for another (marketing, research, predictive modeling). This "function creep" happens gradually and often without clear disclosure to the people whose data is being used.

What CTCs Should Do

Data Minimization: Collect only the data you actually need for legitimate educational purposes. Don't gather information "just in case it's useful later."

Purpose Limitation: Use data only for the specific educational purposes for which it was collected. If you collected attendance data to identify students who need support, don't repurpose it for faculty performance evaluation.

Clear Retention Policies: Establish how long different types of data will be kept and when it will be deleted. Students who graduate or leave shouldn't have their behavioral data retained indefinitely.

Vendor Management: Before contracting with AI service providers, verify their data protection standards. Requirements should be written into contracts, not just assumed.

Transparency: Tell students, faculty, and staff what data is collected, why, and how it's protected. This should be clear, not buried in lengthy privacy policies written in legal language.

The Family Educational Rights and Privacy Act (FERPA) sets baseline requirements for protecting student education records, but many AI tools collect data that falls into gray areas. Err on the side of protecting privacy, not maximizing data collection.

Try This Prompt:

"I'm on a committee developing guidelines for AI tool adoption at our college. What should our data privacy policy include? Give me 5-7 key principles or requirements we should establish to protect student data when using AI tools."

Transparency, Explainability, and Trust

Trust is fundamental to education. AI systems can undermine that trust quickly if they operate as black boxes where decisions feel arbitrary or unexplainable.

The Black Box Problem

Many AI systems can't explain their own decisions, even to their creators. Consider an AI that identifies students at risk of failing a course. Even if accurate, questions remain:

What factors contributed to this assessment?
How can the student improve their standing?
How do we know the system isn't basing predictions on inappropriate factors like demographics?
What recourse does a student have if they believe the assessment is wrong?

Without clear explanations, these systems undermine educational relationships. Students don't trust recommendations they can't understand. Faculty hesitate to use tools they can't explain to students. Administrators worry about liability for decisions made by inscrutable algorithms.

Different Stakeholders Need Different Information

Students need to know when AI affects their educational experience and have options for human review of AI-generated decisions.

Faculty need enough understanding to use AI tools effectively, recognize when outputs are problematic, and maintain appropriate oversight.

Staff need to know how AI tools they use collect and handle data, and when they should seek human backup for complex situations.

Administrators need to understand AI capabilities and limitations well enough to make informed policy and procurement decisions.

Technical staff need detailed information to maintain, troubleshoot, and evaluate systems effectively.

The level of detail varies, but the principle doesn't: people deserve to understand the systems that affect them.

What Makes AI Trustworthy in Education

Trustworthy AI must be:

Reliable: Performs consistently without unexpected failures. If an AI grading tool occasionally crashes and loses student work, it's not reliable. If it performs well in testing but poorly with real student submissions, it's not reliable.

Accountable: Provides clear mechanisms for identifying problems and taking corrective action. When something goes wrong, you need to be able to figure out what happened and who's responsible for fixing it.

Transparent: Operates in ways that relevant stakeholders can understand. This doesn't mean everyone needs to understand the mathematics. It means people can grasp how decisions get made and what factors matter.

Fair: Produces outcomes that align with educational values and equity commitments. This requires ongoing monitoring, not just initial testing.

Respectful of Human Agency: Preserves meaningful opportunities for human judgment in important educational processes. Students, faculty, and staff should be able to override AI recommendations when circumstances warrant.

Try This Prompt:

"An AI system recommended that we [deny admission/flag a student as at-risk/place a student in a lower-level course]. The student is asking why. I don't understand how the AI made this decision. Help me: what questions should I ask to get an explanation I can share with the student?"

Environmental Costs and Sustainability

AI systems consume substantial electricity. Training large AI models requires enormous computational power. Running them at scale (millions of queries per day) adds up quickly.

The Scale of the Problem

Training GPT-3 consumed 1,287 megawatt hours of electricity and generated approximately 552 tons of CO2 equivalent. That's roughly equal to the annual emissions from 120 U.S. homes.

Current estimates suggest large-scale AI systems could potentially reach 102.6 megatons of CO2 equivalent annually if growth continues at current rates.

The Counterargument Deserves Attention

Some researchers argue AI's environmental impact is overstated. They point out:

AI currently represents less than 1% of global carbon emissions
Many data centers increasingly use renewable energy
AI might enable environmental benefits (optimizing energy use, improving climate models, reducing transportation through remote work)
Estimates of water consumption for cooling have been challenged as inflated

The honest assessment: AI does have environmental costs. Those costs are currently small relative to other industries but growing rapidly. Whether the benefits justify the costs depends partly on what we use AI for.

Institutional Responsibility

Higher education institutions increasingly commit to addressing climate change. Many Washington CTCs have sustainability goals. This creates tension when AI tools that could enhance education also contribute to carbon emissions.

CTCs can address this by:

Prioritizing high-value applications: Use AI where it makes significant educational difference, not just for convenience
Choosing vendors wisely: Prefer AI service providers that use renewable energy and publish environmental impact data
Optimizing usage: Minimize unnecessary AI processing (like running multiple checks on the same student work)
Considering alternatives: Sometimes non-AI solutions work fine and use far less energy

The principle: be intentional about AI adoption. The environmental cost is real. Make sure the educational benefit justifies it.

Try This Prompt:

"I want to consider environmental impact when choosing AI tools. What questions should I ask vendors about their energy use and carbon footprint? What alternatives or practices could reduce the environmental cost of using AI in my work?"

Institutional Ethics: Policies and Governance

Individual ethical choices matter, but systemic ethical behavior requires institutional structures. CTCs need comprehensive approaches to AI ethics, not just good intentions.

Governance Structures That Work

AI Ethics Committees: Cross-disciplinary groups (faculty, staff, students, IT, administration) that evaluate AI implementations and develop institutional policies. These shouldn't be IT-only decisions.

Technical Review Boards: Groups with technical expertise who assess AI capabilities, limitations, and security requirements. They answer: can this tool actually do what the vendor claims?

Student Voice: Meaningful inclusion of student perspectives in AI decisions. Students often catch problems others miss because they experience the consequences directly.

Community Input: Engagement with the broader communities CTCs serve. AI decisions affect not just current students but prospective students, employers, and community partners.

Policy Elements That Matter

Acceptable Use Guidelines: Clear standards for when and how AI tools may be used in different contexts (teaching, advising, operations). These should address both faculty/staff use and student use.

Data Governance: Comprehensive approaches to data collection, use, sharing, and protection. This includes vendor management requirements and breach response procedures.

Human Oversight Requirements: Specifications for when human review is required for AI-generated decisions. Higher stakes require more oversight.

Appeals and Remediation: Processes for addressing situations where AI systems cause harm. Students and employees need clear paths to challenge AI-generated decisions.

Regular Review: AI capabilities change rapidly. Policies should be reviewed annually, not written once and forgotten.

The 23% Problem

Only 23% of higher education institutions currently have clear AI policies (EDUCAUSE, 2024). That means 77% of institutions are making AI decisions without comprehensive guidance.

Washington's SBCTC AI Task Force is working to address this gap for our state's CTCs, but policy development takes time. In the meantime, individuals and departments need to make ethical decisions with incomplete guidance.

When official policy doesn't exist yet, apply the AEIOU framework and ask: Is this accessible, equitable, inclusive, open, and universal? If you can't answer yes to all five, proceed carefully or not at all.

Try This Prompt:

"Our college doesn't have an AI policy yet. I'm on a task force developing one. What are the 7-10 essential elements our policy should address? Give me a basic outline we could use as a starting point for discussion."

Individual Responsibility in Institutional Context

Policies provide frameworks, but individuals make daily decisions about AI use. Those decisions have ethical implications whether or not official policy addresses them.

For Faculty

Academic Integrity: Does AI use support student learning or replace it? Are students developing needed skills or becoming dependent on AI assistance?

Fairness: Do all students have equal access to AI tools? Are some students disadvantaged by how AI is integrated into the course?

Transparency: Do students know when and how AI is used in course materials, grading, or feedback?

Professional Development: Are you maintaining expertise needed to effectively oversee AI-assisted processes?

For Staff

Student Support: Does AI enhance human interaction or replace it? Are we using technology to scale support or to avoid personal connection?

Decision-Making: When is it appropriate to rely on AI recommendations for decisions affecting students? When is human judgment essential?

Privacy Protection: Do AI tools handle sensitive information appropriately? Are there alternatives that expose less data?

For Administrators

Resource Allocation: How do we balance efficiency gains from AI with educational quality and equity concerns?

Policy Development: Are we creating policies that enable innovation while protecting students and employees?

Vendor Selection: Are we asking vendors hard questions about bias, privacy, and transparency before signing contracts?

For Students

Critical Evaluation: Does the AI tool you're using actually help you learn, or is it just helping you finish faster? There's a difference.

Academic Honesty: Do you understand your instructor's expectations about AI use? When in doubt, ask before you submit.

Privacy Awareness: What happens to your data when you upload an essay to a free AI tool? Many services use your work to train future models or share data with third parties.

Self-Advocacy: If an AI system flags your work incorrectly or makes unfair predictions about your success, you have the right to challenge it. Document your concerns and ask for human review.

The common thread: ethical AI use requires ongoing attention and judgment. It's not a one-time decision. It's a continuous practice.

Summary

AI ethics in higher education encompasses multiple interconnected dimensions requiring ongoing institutional commitment. The AEIOU Framework (Accessible, Equitable, Inclusive, Open, Universal) provides practical foundation for evaluating AI implementations.

Bias and fairness are critical concerns. Research shows AI systems can produce disparate outcomes across demographic groups, with real consequences for students. Addressing bias requires diverse data, ongoing testing, and meaningful human oversight.

Privacy and surveillance concerns intensify in educational settings where students have limited ability to opt out. Data minimization, purpose limitation, clear retention policies, and careful vendor management are essential.

Transparency and explainability maintain trust in educational relationships. AI systems that operate as black boxes undermine the human connections that make community colleges effective.

Environmental considerations add complexity. AI has real carbon costs, though smaller than some sectors. Institutions should balance educational benefits against environmental impact and choose tools intentionally.

Effective AI ethics requires both institutional governance structures (committees, policies, oversight) and individual responsibility. With only 23% of institutions having clear AI policies, many faculty and staff must make ethical decisions without comprehensive guidance.

Resources

Chan, C.K.Y. (2023). A comprehensive AI policy education framework for university teaching and learning. International Journal of Educational Technology in Higher Education, 20, Article 38. https://educationaltechnologyjournal.springeropen.com/articles/10.1186/s41239-023-00408-3

EDUCAUSE. (2025). AI Ethical Guidelines. EDUCAUSE Library. https://library.educause.edu/resources/2025/6/ai-ethical-guidelines

Future of Privacy Forum. (2024). Vetting Generative AI Tools for Use in Schools. Future of Privacy Forum Policy Brief. https://fpf.org/wp-content/uploads/2024/10/Ed_AI_legal_compliance.pdf_FInal_OCT24.pdf

Gándara, D., Anahideh, H., Ison, M. P., & Picchiarini, L. (2024). Inside the black box: Detecting and mitigating algorithmic bias across racialized groups in college student-success prediction. AERA Open, 10. https://journals.sagepub.com/doi/10.1177/23328584241258741

Gutierrez, J. (2025). AEIOU Ethos: A Framework for Responsible AI. KDP. https://www.amazon.com/AEIOU-Ethos-Framework-Responsible-AI/dp/B0DVJGNLS3

Khosravi, H., Shum, S. B., Chen, G., Conati, C., Tsai, Y. S., Kay, J., Knight, S., Martinez-Maldonado, R., Sadiq, S., & Gašević, D. (2022). Explainable artificial intelligence in education. Computers and Education: Artificial Intelligence, 3, 100074. https://www.sciencedirect.com/science/article/pii/S2666920X22000297

MIT News. (2025, January 17). Explained: Generative AI's environmental impact. MIT News. https://news.mit.edu/2025/explained-generative-ai-environmental-impact-0117

Nazaretsky, T., Ariely, M., Cukurova, M., & Alexandron, G. (2025). The critical role of trust in adopting AI-powered educational technology for learning: An instrument for measuring student perceptions. Computers and Education: Artificial Intelligence, 6, 100213. https://www.sciencedirect.com/science/article/pii/S2666920X25000086

Stanford University. (2025, January). Report outlines Stanford principles for use of AI. Stanford Report. https://news.stanford.edu/stories/2025/01/report-outlines-stanford-principles-for-use-of-ai

U.S. Department of Education, Office of Educational Technology. (2023). Artificial Intelligence and the Future of Teaching and Learning: Insights and Recommendations. https://files.eric.ed.gov/fulltext/ED631097.pdf

Vassel, F.-M., Shieh, E., Sugimoto, C., & Monroe-White, T. (2024). The psychosocial impact of Gen AI harms. AAAI Spring Symposium. Stanford HAI. https://hai.stanford.edu/news/how-harmful-are-ais-biases-on-diverse-student-populations

Yu, Y., Wang, J., Liu, Y., et al. (2024). Revisit the environmental impact of artificial intelligence: the overlooked carbon emission source? Frontiers of Environmental Science & Engineering, 18, 132. https://link.springer.com/article/10.1007/s11783-024-1918-y

When you've finished reviewing the material, click 'Next' to check your knowledge.