Responsible Tech: Generative Images in Education - Workshop Results

Responsible AI continues to develop alongside new uses of AI in education. At this years’ onderwijsdagen, Duuk Baten and I took the opportunity to challenge our audience and support them in shaping a better and more responsible future.

The workshop, Immersive ethics, walked participants through three different questions.

  • Which value is the most important for the future of Gen AI in education?
  • Can you develop/derive questions from the public value you chose?
  • What should we do about your concerns/questions?

In a nutshell, the concerns of our participants highlight the need for clear guidelines, regulations, and ethical frameworks to address the challenges associated with the use of AI image generation tools in education, such as midjourney.

A midjourney created image using the prompt:  ‘Create a user worried looking at a computer neural network’
A midjourney created image using the prompt: ‘Create a user worried looking at a computer neural network’

The common topic – Midjourney

Midjourney, a popular AI image generator, was used as a focusing example for the workshop. The tool allows for users to create images, often seemingly from nothing, to use to their liking. Midjourney generates images from natural language descriptions, called prompts, similar to OpenAI's DALL-E and Stability AI's Stable Diffusion. (Wikipedia, 2023)

For education, midjourney has showed potentially new ways of shaping education  in use such as:

  • Creative Exploration in education and research
  • On-the-fly generation of visual content during
    lessons or presentations
  • Future visions, reimaging historical events
  • Speed and ease of use

More importantly, real challenges have also been presented that raise new ethical and social obstacles to using such a tool. 

  • Reinforcing Biases
  • Originality: Ownership Dilemma
  • Plagiarism: Detection Challenges
  • Deepfake creation and curation
  • Emergent risks in political use cases

Through the workshop we found that public values helped discuss and shape meaningful conversation while still moving towards shared actions .

Most common values listed
The most common values raised during the workshop

Overview of values and the concerns that come with them

Transparency, Ownership, and Privacy:

  • Exploring the need for transparency in AI image generation and understanding data usage
    • Participants discussed how the need for transparency continues to increase from tools such as Midjourney. Where do image data come from? How can I reproduce my creation? Where do prompts go after they are sent to the server?  


  • Addressing ownership rights and implications for users' creative outputs
    • Throughout the workshop participants discussed who owns what. Are learners allowed to use AI images? Do they own the image? Where will we store all of these new images learner’s and educators produce?


  • Examining privacy and data protection concerns and ensuring compliance with regulations
    • Compliance with copywrite laws and regulations continues to be challenged as generative ai image tools popularize. With the AI act coming in the near future, “Companies deploying generative AI tools, such as ChatGPT or image generator Midjourney, will also have to disclose any copyrighted material used to develop their systems.” (Reuters, 2023)
      • What should we do?
        • Participants wanted assurance that the authorship rights of AI-generated images are trustworthy, and that the authenticity of these images can be guaranteed. They also want to protect copyright and prevent unauthorized use or manipulation of these images. Regarding privacy, participants want to know if the same laws apply to these images as they do to other forms of personal data.

Trustworthiness, Authenticity, and Ethical Use:

  • Highlighting the importance of establishing trust in AI-generated images and their authenticity
    • Similar to transparency, trust and authenticity must be acknowledged as leading to better more valuable outcomes from these tools. This includes trust between users, where learners, educators and institutions alike should be honest when using tools like Midjourney for their work. Citing, referencing and discussion about the images should be commonplace.


  • Discussing ethical considerations and potential influence on students' learning experiences
    • Learners must be asked new questions about apps like Midjourney. How will students’ work with these applications? Will they use these apps in their own presentations, talks, or study? Which students have access to these tools and is it important to their learning style? More questions can be asked, but the conversation needs to happen even sooner.


  • Promoting responsible use through education, awareness programs, and guidelines
    • Concerns: Within that conversation, participants acknowledged the need for more awareness, agreements and action. Participants were concerned about the potential misuse of AI-generated images, particularly in educational settings. They want to prevent conscious or unconscious influence by AI and ensure that students understand the limitations and potential ethical implications of using such tools.
      • What should we do?
        • Participants emphasised explicitly including ethics in policies related to AI usage in schools. This suggests a concern for ensuring responsible and ethical use of AI technology. Their responses highlight the importance of getting students actively involved and moving in the learning process. For instance, this could include creating interactive and engaging educational experiences using AI tools in a learning environment.
Anne Fehres and Luke Conroy & AI4Media / Better Images of AI / Hidden Labour of Internet Browsing / CC-BY 4.0
Anne Fehres and Luke Conroy & AI4Media / Better Images of AI / Hidden Labour of Internet Browsing / CC-BY 4.0

What else can we do?

Other topics came up as well at this workshop. The following are more advice and ideas shared by the group.

  • Raising awareness about the possibilities and limitations of AI in education: For this insight is necessary on the origin of of generative AI use as well as what data this based upon. However, considering the dynamic impact of AI in education, participants mentioned the importance of following the many eyes principle towards dealing with risks – to be able to pragmatically go forwards. They also mentioned the need to avoid prejudice, indicating a focus on promoting critical thinking and unbiased perspectives.


  • Citizenship and social cohesion: The idea of including responsible citizenship as a subject in the curriculum was considered, highlighting the possibilities of understanding AI education and it’s impact on future citizenship and how these tools can change our social cohesion norms.


  • Collaboration and multidisciplinary teams: There was also a suggestion of setting up multidisciplinary teams within our institutions to create broad awareness. This is recognition for the need for more collaboration and diverse expertise when implementing AI in education for the future.


  • Reliability: The outputs of models like Midjourney are not reliable or easily repeatable. For education, this raise concerns that should be addressed through new ways of experimentation. One suggestion is controlled pilots that are conducted in real educational settings, utilizing various types of LLMs owned by both private and public entities. These pilots would aim to challenge organizational values and provide valuable insights into the advantages and disadvantages of integrating LLM-based tools into the classroom.

The greater context and taking action

The Immersive Ethics workshop was more than just a discussion - it was a call to action. It challenged us to think critically about approaches to Generative AI in education and reminded us that at the heart of technological progress must always be a commitment to responsible, ethical practices.

Overall, our workshop reflected the thoughtful and comprehensive need for public values in incorporating generative image AI in education. From the workshop we see a desire from educational institutions to have a focus on greater AI ethics awareness, student engagement, data security, reliability of models, responsible citizenship, and collaboration.

We now hand off the results of this workshop to the Data and AI pilothub by the Npuls (Find out more). These results will help better the pilot’s vision and understanding of our member’s needs when it comes to generative AI in education.

SURF and Npuls continue to try to better understand the questions you have about generative AI and have created a survey to capture them. Please take a minute to add yours here (link to vraagbaak-survey)!

Interested in responsible tech? Check out our new discussion paper and send us your thoughts about recommendations, tools and insights you gather from your own responsible practices.



Dit artikel heeft 0 reacties