The Role of Responsible AI
in Higher Education
Dr. Nava Shaked
School of Multidisciplinary studies, HIT, Israel
13th International week| Chania 2026
1
🎓
2
“A student just submitted their best paper ever.
Did they write it?”
Does it matter HOW?
Does it matter WHAT they learned?
Or does it matter WHO they became?
These questions define the challenge of Responsible AI in education today.
🤔
Watch & Think
3
Three short clips (<60 sec each) — scan the QR code or open the link on your phone
😂
FUNNY
“Did You Use AI?”
Teacher vs Student — the classroom meme that went viral in 2025
🚨
TROUBLING
Falsely Accused by AI
University uses AI to accuse thousands of students of cheating — wrongly (ABC News 2025)
youtube.com/shorts/YahFFxTx2g0
🧠
PROVOCATIVE
AI Grading Bias
How AI grading algorithms unfairly judge students — bias baked in from day one
🎬
AI Is Already in Your Classroom
4
92%
of UK students use GenAI tools for learning
60M+
students & teachers on Quizlet worldwide
18M+
ELSA Speak downloads in 100+ countries
>100M
students exposed to GenAI by end of 2024
Sources: WEF Education 4.0 2024 Quizlet ELSA Speak UNESCO 2023
📊
The Tools Your Students Are Already Using
5
ChatGPT
Text generation, research, assistant, coding
Brisk Teaching
AI feedback & lesson planning for teachers
Quizlet
AI-powered adaptive, flashcards & study guides
ELSA Speak
AI pronunciation & English speaking coach
Grammarly
Writing improvement, tone analysis, citations
Perplexity
AI search engine with cited sources
🛠️
Our 25-Minute Journey Together
6
01 The Reality Check AI is already here — let the numbers speak
02 What Responsible AI Means Six pillars, real frameworks, global standards
03 Real Tools, Real Tensions Three case studies: ELSA, Brisk, Quizlet
04 Rethinking Pedagogy What changes when AI can explain anything?
05 From Talk to Action A roadmap institutions can start using today
📋
AI Is Not Neutral — And That Is Our Problem
7
Algorithmic Bias
61% of ESL essays wrongly flagged as AI-generated. Proctoring tools fail Black students 57% of the time.
Hallucination Risk
Quizlet AI generated a confident study guide about a Nazi WWII victory — completely fabricated history.
Privacy & Surveillance
EU AI Act bans real-time emotion detection and biometric monitoring in classrooms.
Equity Gap
2.7 billion people lack reliable internet. Personalization only works for the already-privileged.
⚠️
The Rules Are Being Written Now
8
UNACCEPTABLE — BANNED Emotion detection, Biometric monitoring, Real-time facial recognition for discipline
HIGH RISK — Strict Compliance AI grading, admissions tools, dropout prediction, binding tutoring systems
LIMITED RISK — Transparency Required AI chatbots (disclose as AI), writing assistants, recommendation engines
MINIMAL RISK — Deploy Freely AI-enabled games, Spam filters, Language practice tools
EU AI Act: In force Aug 2024, Fully applicable Aug 2026, Penalties up to 35M euros
⚖️
What “Responsible” Actually Means: 6 Pillars
9
Fairness
Outcomes must not discriminate by race, gender, language or socioeconomic status.
Transparency
Educators and students must understand how AI decisions are made.
Privacy
Student data: collected minimally, stored securely, never repurposed without consent.
Accountability
AI augments — but humans remain answerable for every decision.
Safety & Well-being
AI must not harm students emotionally, socially or academically.
Inclusiveness
AI must serve ALL learners — different needs, languages, cultural backgrounds.
🏛️
Real Tool, Real Tensions: ELSA Speak
10
18M+ downloads | 100+ countries | AI-powered English pronunciation coach | Free tier available
Transparency Gap
How does ELSA score pronunciation? Which accent is the “correct” standard? American? British? The model is a black box.
Fairness Risk
Research shows AI language tools systematically disadvantage non-native speakers. An accent is not a mistake — but ELSA may treat it as one.
Privacy Concern
Voice biometric data retained “as long as deemed necessary.” May be used to improve AI. Multiple third-party access points.
Key insight: Strong regulatory compliance does NOT equal full Responsible AI. (Analysis: Morgenbesser & Marom, 2025)
🎤
When AI Gets Teaching Right: Brisk
11
Free Chrome extension | Integrates with Google Docs, Slides, PDFs | 30+ AI tools for teachers
Human-in-the-Loop
Brisk consistently reminds teachers to review AI content BEFORE sharing with students. AI is advisory — human judgment is final.
Safety First
Built-in filters block harmful content. If a student uses abusive language, the teacher is notified immediately — automatic escalation.
Privacy by Design
FERPA, COPPA and GDPR compliant. Student data is NOT retained after tasks complete. No third-party tracking. No AI training on student inputs.
Takeaway: Brisk shows that RAI design is not a constraint on usefulness — it is what makes a tool trustworthy. (Analysis: Rudzki & Erez, 2025)
⚡
When AI Gets It Wrong: Quizlet
12
60M+ users worldwide | AI-powered study platform | Generative AI for flashcards and study guides
The Hallucination Incident
A student asked Quizlet’s AI to create a study guide about “the Nazi victory in WWII.”
The AI generated a confident, detailed plan — based entirely on fabricated history. No flag. No disclaimer. No refusal.
Task completion was prioritized over factual accuracy and safety.
RAI Principles Violated
Transparency: AI presented fiction as fact with full confidence.
Safety: No guardrails for historically harmful content.
Accountability: Who is responsible when AI poisons student knowledge?
Lesson: Popularity and scale do not guarantee responsibility. Without safety guardrails, AI at scale is a trust liability. (Analysis: Shor & Cohen, 2025)
🧠
Responsible AI Is an Ecosystem, Not a Checkbox
13
Governance
Accountability structures, ethical review boards, risk tiering, impact assessments, decision-making frameworks.
Engineering
Fairness toolkits, robustness testing, explainability (XAI), adversarial drills, bias detection in CI/CD.
Research
New methodologies for safety, fairness, interpretability, and human-AI collaboration.
Policy
Legal compliance with GDPR, EU AI Act, OECD principles. Alignment with global AI regulations.
“RAI is not a destination. It is a living ecosystem that evolves with society and technology.”
🌐
14
What if the real disruptor
isn’t AI?
It’s us?
AI enters through our choices: which tools we deploy, which values we embed, and what we choose to do with what we know.
💭
The Educator’s Role Has Fundamentally Changed
15
FROM
TO
If AI can explain almost anything.
The teacher’s value is now in what AI cannot do.
🔄
The Question That Changes Everything
16
What does learning mean
when knowledge is free?
Skills shift
From storage to evaluation, synthesis, and judgment
Assessment shifts
From “what do you know” to “what can you do with what you know”
Credentials shift
Degrees must signal capability, not just completion
❓
Skills That Cannot Be Downloaded
17
Jacob Morgan’s Future Leader (2020) — reframed for the Responsible AI era
Explorer
Always learning, embracing curiosity, asking questions AI cannot formulate. Drives innovation through human wonder.
Chef
Balancing technology with human creativity and judgment. Knows when to use AI and when to set it aside.
Empowerer
Using AI to uplift others and promote inclusion. Leading for human flourishing, not just efficiency.
Global Citizen
Navigating cultural complexity, ethical ambiguity, and diverse perspectives that AI cannot fully understand.
🚀
A Day in 2030: The Responsible AI-Powered Campus
18
Three insights from a near future where Responsible AI is the institutional norm…
The Student
Maya logs in. Her AI tutor has reviewed her draft and flagged two areas of weak reasoning — not grammar, not style. Her instructor sees the same report and adds a voice comment. AI and human, working together.
The Instructor
Prof. Lior receives an algorithmic alert: three students are at risk of disengaging. He reviews the data, decides to restructure next week’s seminar. The AI suggested — he decided.
The Institution
The university’s RAI Board publishes its quarterly AI impact report: which tools are deployed, what data they use, what oversight exists. It is publicly accessible. Accountability made visible.
🔮
From Talk to Action: 5 Steps for Your Institution
19
01 Audit: Assess ALL current AI use — formal and informal. Map which tools, who uses them, and whether any policy governs them.
02 Engage: Bring faculty, students, IT and leadership into the conversation. RAI decisions cannot be made by one department alone.
03 Define Targets: Set realistic 6-12 month objectives. Which pillar needs the most work? Bias? Transparency? Data governance?
04 Create Policy: Develop ethical guidelines and usage standards. Reference the EU AI Act, UNESCO frameworks, and OECD principles.
05 Build for Scale: Design policies and tools that can grow. RAI is a journey — build systems that evolve as AI evolves.
🗺️
The Question That Defines Our Era
20
The question is not whether AI belongs in higher education.
It is already here. The question is
who & how to shape it.
Ignore it
It will shape your students without you.
Fear it
And miss the opportunity to guide it wisely.
Lead it
Build the future of learning, responsibly.
✨
Thank You — Let’s Think Together
21
That is what responsible AI education looks like.
Questions?
Dr. Nava Shaked
School of Multidisciplinary Studies
Holon Institute of Technology HIT, Israel
Shakedn@hit.ac.il
“It is up to us.”
👋