Building on the previous discussions from this toolkit, GenAI tools bring both opportunities and challenges to teaching and learning contexts. Apart from acknowledging the pitfalls of GenAI, also referred to as the uglies, this module aims to support educators with critical reflection tools and strategic responses to address the less desirable aspects of GenAI.

First key questions:

• What specific risks do you see GenAI introducing to your teaching context?
• Can you identify potential unintended consequences of GenAI that are immediately obvious?
• Describe a specific instance where GenAI tools created challenges in your context. What strategies did you develop to address these challenges?

Suggestive Responses to Key Risks/Challenges

(1) Bias reinforcement

Platforms perpetuate biases from the datasets they are trained on (Kooki, 2023). User interactions can potentially amplify existing stereotypes and discriminatory practices.

Strategic responses

• Continuously monitor AI-generated content for bias.
• Educate ourselves and learners about potential biases.
• Encourage developers to center users’ diverse social identities and develop culturally responsive AI deployment strategies (Kannan, 2024).

(2) Misinformation

Potential spread of inaccurate or unverified information challenges our ability to distinguish credible sources.

Strategic responses

• Integrate information literacy modules into the course, especially when allowing students to use GenAI in the course assignments.
• Teach cross-referencing techniques.
• Teach proper citation and academic integrity.

(3) Overreliance

Students become overly dependent on GenAI tools, which potentially reduces critical thinking skills and engagement in social learning processes

Strategic responses

• Create AI detox learning experiences (No AI Day, Activity: what is my life now without AI?)
• Develop assignments that challenge students to work without AI engagement to encourage original thought and complex problem-solving.

(4) Data privacy

Platforms with unclear data collection and usage policies often require access to user data to personalize and enhance their services. In the context of teaching, they may collect personal data including students’ names, grades, learning patterns, personal interests, social interactions, etc. When the AI tool is granted access to this information, there is a high risk of data breaches and misuse.

Strategic responses

• Thoroughly review and understand platform-specific data policies.
• Teach students about data privacy and how to protect themselves.
• Advocate for transparent consent mechanisms.
• Choose tools with strong privacy protections.
• Learn to use local AI.
  • ChatGPT (by OpenAI)
    This platform trains on user data by default everywhere except Europe. However, you can choose to opt out as outlined below.
    • Settings > Data Controls > Improve the model for everyone. Turning it off applies only to the future.
    • When you use the temporary chat, your data will not be used for training, regardless of whether you opted out.
  • Claude (by Anthropic)
    • Anthropic doesn’t train on user data for any of their products, individual or commercial. The three exceptions are when the content is flagged due to a trust and safety issue, the user explicitly reports the content, or the user explicitly opts into training.

(5) Environmental dilemma

The development, maintenance, and disposal of AI technology comes with a large carbon footprint, electronic waste, and impacts on natural ecosystems.

Strategic responses

Here are some ways to reduce your environmental footprint as inspired by Emily Simpson’s practical tips to minimize your carbon/water usage from AI that were shared in the presentation Climate Conscious AI Use: Wrestling with Environmental Impacts at ETUG Fall 2024.

• Choose the right tool for the task. For example, use smaller, less resource-intensive models for simpler queries or set up a custom chatbot with a smaller AI model for common, simple tasks.
• Recycle and reduce: reuse previous AI-generated outputs to save unnecessary re-computation or encourage group work or group demonstrations to minimize computation requests.
• Limit output length: reduce the computational effort/energy by being intentional and precise in the original prompt to tailor output.
• Group multiple questions or tasks into a single request to reduce computational resources.
  • For example, please complete the following tasks, restating each prompt before providing the answer:
    • Task 1:                                                                                                 
    • Task 2:                                                                                                 
• Run a local model on your device as a small local model does not use cloud storage or communication, and consumes less energy than applications like ChatGPT.
• Clearly communicate the values of sustainability and transparency about the AI-related environmental impacts of the product, especially if you are part of the team responsible for deciding whether to purchase the AI tool for the department or institution.

(6) Depersonalization of learning

A decrease in human interaction reduces the development of social skills and the emotional intelligence of students.

Strategic responses

• Open spaces for discussion related to the erosion of human connection when overusing AI in daily life.
• Enhance human collaboration and relationships by designing learning experiences with GenAI.
• Extend educators’ roles to include providing emotional support and mentorship and fostering a sense of community, which GenAI cannot fully replicate.

(7) Accessibility and digital divide

Disparity in access to technology can lead to a digital divide, exacerbating existing educational inequalities and creating a divide between those who can benefit from AI-enhanced learning and those who cannot.

Strategic responses

• Discuss the importance of designing or integrating AI tools that are accessible and inclusive to all potential users. For example, providing text-to-speech for visually impaired individuals or predictive text for those with motor impairments (acknowledged and managed).
• Be mindful of the digital divide when integrating technologies in teaching.
• Be ready to offer continuous support when engaging with AI platforms.

Further Readings

• Ifelebuegu, A. O., Kulume, P., & Cherukut, P. (2023). Chatbots and AI in education (AIEd) tools: The good, the bad, and the ugly. Journal of Applied Learning & Teaching, 6(2), pp. 332-345, https://doi.org/10.37074/jalt.2023.6.2.29
• Kannan, P. (2024). How harmful Are AI’s biases on diverse student populations? Standford University: Human-Centre Artificial Intelligence

Summary

This module critically examined the complex landscape of GenAI in teaching and learning in higher education settings, acknowledging that these tools come with notable challenges. Recognizing the potential uglies or pitfalls of GenAI technologies, the module was designed to support educators with essential reflective tools and suggested proactive strategies. The holistic approach addresses not only technological concerns, such as bias, misinformation, and data privacy, but also deeper pedagogical implications including the potential erosion of human connection, critical thinking, and educational equity.

Final Reflective Question

How have you personally responded to one of the challenges posed by AI in your educational practice?

What would be the next steps you take in response to some of the challenges posed by GenAI in education?