Recently, I had a student email my co-lecturer with a concern about a class that I was conducting. This student refused to use a Gen AI image creation tool that I was trying to use for the class. The concern and rationale was that these AI tools steal from real artists and also diminish patient experiences in healthcare. As a digital health practitioner, this email struck a personal chord. As I reflected on my role as a digital health practitioner, an educator, and a researcher — it made me realize that the field of digital health is at a critical crossroads right now. The recent news just over the Chinese New Year on new Gen AI bots — DeepSeek and Qwen — has also disrupted the AI world. In fact, the news about DeepSeek is evolving so fast, and many countries and organizations are concerned about its privacy, confidentiality and cybersecurity, since DeepSeek was from China. The ethical and legal challenges of Gen AI are now more important than ever. Hence, I have decided to write this post to you, my fellow practitioners and learners in digital health and education, to share some of my personal thoughts on how to embrace Gen AI responsibly in our healthcare practices, education and innovations. There are many aspects to address, and if I talk about all of them, this post will be never-ending. Therefore, I shall focus on the ethical use of AI Art here, and perhaps address other concerns in future posts.

Gen AI Art Creation Tools: Boon or Bane?

There are many AI art generation tools — some popular ones are DALL-E, Midjourney, DreamStudio, DeepAI, Leonardo AI, among others. Even traditional image websites like Shutterstock, Getty Images, Canva and Adobe Firefly are incorporating Gen AI into their applications. In fact, the popular Gen AI chatbots like ChatGPT and even Bing Image Creator are able to generate images within the bots as well. So far, my favorites are Dream by WOMBO and NightCafe.

These AI art tools have transformed creative expression by producing detailed, intricate images and pictures through user inputs. Nowadays, you can ask an AI art tool to create an image of almost anything, all just from a text prompt! By choosing from various AI models and the types of images that you want (e.g. photorealistic, cinematic, anime, etc), one can create an impressive piece of “art” from these tools. Trained on huge datasets of artworks, which may or may not be copyrighted, and with or without the explicit consent of the original artists, the fact is that AI art generation tools are probably going to “democratize art” even further as technology advances. Yes, there are controversies surrounding these AI artworks, but in my opinion, the key lies on users like us to use these tools in an ethical manner — as the tools themselves are of not much use without the user’s creativity and ethical personality. Let me first list the questions, controversies and concerns surrounding these AI art.

Key Concerns Regarding IP Rights and Creative Ownership:

• Training Data and Artists’ Consent: There have been criticisms about how the training of billions of artworks available onlnine essentially “steals” the unique styles of artists, as many of these works were used without the original artists’ permissions or compensations. One of the most famous legal cases is the one on Andersen v. Stability AI, which highlighted the contention between “art” created by AI-generators and those of the original artist. A copyright infringement lawsuit was filed against several AI art companies by the copyright owners — the claim being that these AI-generated images can be derivative works of the original artists, and thus infringed on the original copyrights. In 2023, a lawsuit also made the headlines between Getty Images and Stability AI, as the stock photo provider accused the AI-generator company of misusing more than 12 million of its photos to train their AI-image model. In the midst of ongoing debate about this, it is clear that AI-generated “artworks” have already “democratized art” in a way that was never expected. It will evolve as technology advances, but it will never go away. So how should we balance our ethical conundrum in this area as AI and digital health practitioners?
• Original and Derivative Works: As we continue to live in co-existence with AI technologies, the line between replicated/duplicated works and creativity/inspiration becomes blurred even further. In this regard, the US Copyright Office has recently introduced new guidance, which states that purely AI-generated works that lack significant human creativity, are not eligible for copyright. It is clear that human creativity and control in the creation process is essential in the final “artwork” in order for copyright to be considered – the AI-generated content must have “sufficient human authorship” and just “expressive elements” generated through user prompts may not necessarily quality for copyright protection. The US Copyright Office recognizes the complex interplay that exists between AI-generated works and human creativity, thus they stress the need to provide transparency in any works submitted in the copyright application process, including both the AI tools used and the AI-creation process. The new guidelines advocate that in relation to “copyrightability”, at its current stage, regardless of the length of the prompts, since the final AI ouput is the user’s acceptance of the AI system’s interpretation, the final output is not copyrightable. In summary, the new guidelines sets a difference between: (i) AI used as a tool to assist in creative work; versus (ii) “AI as a stand-in for human creativity”. If the author is “referencing” and not “incorporating the final output” (e.g. outlining a book, coming up with song ideas, or artists using the AI to modify their own works), the final human-produced work may still be copyrightable, depending on a case-by-case basis. In short, the AI-generated elements are not copyrighted, but the “perceptible human expression” in the work would still be copyrightable.

“No matter how many times a prompt is revised and resubmitted, the final output reflects the user’s acceptance of the AI system’s interpretation, rather than authorship of the expression it contains.”
– By US Copyright Office Report: Copyright and Artificial Intelligence: Part 2: Copyrightability

Our Roles as Digital Health Practitioners & Educators

As an advocate of digital health both in my research and practice, and as a passionate educator, I believe that Gen AI literacy is needed for students and learners to understand the pros, cons and grey areas of this rapidly evolving field right now. As a Certified Prompt Engineer, I strongly believe that good quality prompting is the way to go to effectively use Gen AI to supplement and/or complement the work that we do. We should not shun away from AI technology, but rather explore how we can use it as a partner in our innovations, rather than a replacement of our own creativity. After all, creativity is still what makes us human…

Ethics of Using Gen AI-based Art for Healthcare Experiences

In healthcare, Gen AI is revolutionizing clinical practices and clinical education. AI-generated images can not only enhance clinical training, but also support research and even educate patients and healthcare learners. However, the student’s concern about AI-images diminishing patient experiences, is a valid one. Hence, let me try to address this ethical dilemma and concern.

Can AI Protect the Authenticity and Digity of Patients’ Experiences?

Coming from a humanistic perspective, there is a real fear that AI-based images can potentially play down on patients’ experiences to just data points, making them either impersonal, dehumanized or detached from reality. These concerns are definitely real. Does AI art destroy the dignity of patient’s experiences, or can it potentially preserve the authenticity of such experiences? Let’s look at this issue from both sides of the coin…

Commodification of Experiences Due to Bias and Misrepresentation

There were criticisms regarding the biasness of training data for Gen AI when it first came out — ranging from stereotypical biases resulting from certain genders, races/ethnic groups, cultures, jobs, and even ages. While these biases may still exist, the fact that users all over the world are now “feeding” Gen AIs with data through their prompts will no doubt lead to an evolution of the algorithms that are working behind these AI systems. Suggestions to reduce these biases in Gen AI include the use of diverse datasets to train newer models, ensure robust testings of the Gen AI applications (such as checking the AI-generated answers to ensure that the information is accurate, reliable, relevant and appropriate), ensure transparency in the decision-making process behind AI algorithms, and conduct regular updates and monitoring of the AI model and the datasets. Many AI ethical guidances have been developed by various international organizations, such as the European Commission, US Agency for International Development, UNESCO, World Economic Forum, and World Health Organization. Locally in Singapore, AI guidances have also been developed by the Personal Data Protection Comission (PDPC) and Infocomm Media Development Authority (IMDA). Specific to Singapore’s healthcare, an Artificial Intelligence in Healthcare Guidelines (AIHGle) was also developed by the local health authorities. Although none of these specifically addresses AI-generated “artworks”, the underlying principles are still the same. Going through all these guidances in this blog post is not possible, but perhaps I will consider discussing these guidances in a future class of mine.

Healthcare Professionals vs Gen AI: Is the World Big Enough For Both to Co-exist?

As healthcare professionals at the workplace, we know that as technology advances, AI (and Gen AI) will potentially be integrated into our daily practices. While the ethical dilemmas above are very real, this pushback and resistance against Gen AI may have potentially come from a place of fear — a fear that it may undermine the authenticity of our work and care of patients, or a fear that it may take potentially take over our jobs. However, just simply rejecting AI/Gen AI will only leave us ill-prepared as healthcare professionals in an increasingly AI-world. Thus, as practitioners and educators, we need to start thinking about how we can engage meaningfully with these Gen AI systems both critically and thoughtfully.

Interestingly, a Vatican document just released a few days ago has advocated that there needs to more human responsibility as the world becomes more technologically advanced. While the document states that AI can only “simulate” the authenticity of human relationships, it acknowledges that some people may seek AI in search for meaning and fulfillment as well. In fact, in a landmark study that was published in 2023 to identify if Gen AI chatbots (ChatGPT) could provide better and more empathetic responses than healthcare professionals, their results showed the participants (who, by the way, were all healthcare professionals) had identified ChatGPT’s responses as being more empathetic and of better quality!

I am sure more studies like this will surface in the near future. How then should we approach Gen AI? As AI gets more advanced and more powerful, in order to use these AI tools responsibly, we need to understand the benefits and limitations of these technologies, but also embrace their integration into our practices with critical insight. As educators, we should not only help our students and learners increase their AI literacy, but more importantly, enhance their curiosity and critical thinking skills, so that they will be able to face and navigate this “unknown beast” with confidence, yet caution, in their future practices. Although I am a fervent digital health advocate and practitioner, I totally agree with the part of the Vatican document that emphasizes that AI is not a substitute for “authentic human relationships” and should be used to enhance, but not replace the patient-practitioner relationship.

Leveraging Gen AI to Preserve Patient Narratives

Although critics may worry about the depersonalization of patient experiences thorugh AI-generated images, if you look at these technologies from the other side of the coin, these features can also be used (in an appropriate manner) to empower patients by providing a creative outlet to express their feelings about their medical conditions, medications, and treatments.

In my own digital health research, such as MAGIC (Metaverse Art Gallery of Image Chronicles), I have experienced firsthand how Gen AI art-based images of hero-villain portraits can be used positively to reflect patients’ lived experiences of their medication therapies. These superhero and villain representations of medications help patients articulate their experiences in a creative, engaging and safe manner. Of course, in order to ensure that the AI-based image is accurate, reflective and symbolizes the patient’s lived experience, it is crucial that we use an evidence-based framework to develop these portraits — and to this end, we collaborated with our hospital’s Senior Art Therapist to work with the patients to co-design their portraits, so as to ensure that the resultant AI-images are both accurate and appropriate. Rather than diminishing the human element, these images can empower patients by protecting their privacy, yet providing a novel medium to express their complex emotions.

Enhanced Creative Expression through AI-powered Art Therapy

I had previously written a post explaining what is Art Therapy. Many of us would have the misunderstanding that Art Therapy is just about drawing pictures or making art objects, but working with our Art Therapist, I have come to realize that this is definitely not the case! Art Therapy is different from Art as Therapy (see Figure below), and uses special psychotherapy techniques to help patients express their innermost thoughts and feelings beyond the use of language and text, through art. By exploring these feelings, patients can experience an increased sense of self-awareness and a form of emotional release that traditional medication therapies cannot fill.

Besides the work that our group has done, many other groups overseas are also exploring Gen AI-based art for patient care — an example is CareYaya, who is backed by Harvard Innovation Labs. They are currently exploring how AI-powered art therapy can help hospice patients. Universities like Yale School of Medicine are also exploring how AI-created art can help improve student education, in terms of enhancing narrative medicine and creating educational imageries. As part of their curriculum, medical students are taught to reflect and share their insights into patient or clinical experiences through AI-generated art — providing new ways of health communication, creative expression and reflection. In terms of patient care, AI-based art therapy can also empower patients to engage in the creative process, thus providing a new avenue for comfort and healing for their hospice patients. Furthermore, AI-based art therapy can make an impact on certain groups of patients who are not able to participate in traditional art therapy sessions, for example, elderly with Parkinson’s, or patients with life-limiting diseases and/or disabilities. Have a look at the videos below on how AI-powered art therapy is changing the healthcare landscape.

Advantages of using AI-powered Art Therapy

By combining the evidence-based concepts of art therapy and the technological advances of Gen AI, patients’ lived experiences can be tailored through a sense of empowerment and autonomy in their creativity. They can find new methods of self-expresison and emotional release that was not previously possible with traditional art therapy techniques. Through AI technologies, patients can have a deeper sense of connection with themselves and others, as well as have a renewed sense of emotional well-being and comfort.

For certain groups of patients, Gen AI chatbot and virtual companions can offer a sense of emotional support and help improve patients’ feelings of loneliness and mental well-being. For ideas that are abstract, and cannot be easily depicted through manual drawings, stock photos or Google searches — Gen AI can potentially help in visualizations and enhance the understanding between the patient and healthcare professional.

In AI-based art therapy, a reflective process is used in the image creation process. When the pictures are generated, the patient or user can modify the words or use more variations to generate different images that convey a more personalized and accurate meaning. This iterative process helps the user crystallize their narratives into something that is symbolic and representative, thus potentially making the communication process between the healthcare professional and patient more effective.

Besides saving on physical art materials and consumables, educational resources — such as medical illustrations/diagrams or patient education materials/ visual aids — can be generated through AI art to convey the necessary information to students about the lessons, or patients about their diseases and medications, without infringing the copyright of the images found online.

But perhaps the most crucial advantage of Gen AI art in healthcare education, is the fact that it can enable the sharing of patients’ experiences with students without the fear of invading the privacy and confidentiality of the patient. These art creations can convey the patient’s narrative without violating health privacy and security concerns when teaching, yet complying to international and local regulations, such as the HIPAA (Health Insurance Portability and Accountability Act of 1996) and PDPA (Personal Data Protection Act), since an image that accurately represents the patient’s case can be generated without the use of real patient photographs. This is especially useful in healthcare and medical education, since students and learners need to gain some form of practical experience, yet preserving patients’ confidentiality is of utmost importance.

Final Thoughts

As healthcare professionals working in the age of advanced technology, we stand at the crossroads where the transformative potential of Gen AI is shrouded with the uncertainties of authenticity of patient experiences, potential AI bias and creative ownership of human-computer interactions.

We know that AI is here to stay, and it will continue to evolve rapidly with time. Instead of avoiding it out of fear, we must be proactive users to become more informed, so as to ensure that AI will serve clinical practice and patient care in ethical and meaningful ways. How AI and Gen AI will evolve in healthcare depends on how we integrate and co-exist with these technologies in our practices and education. As an interdisciplinary digital health practitioner, I believe that with the right balance of “art”, “science” and “tech”, we can effectively harness AI’s potential in healthcare, yet protect patients’ rights, preserve the humanity of patient experiences, and foster responsible innovations in healthcare practice and education.

In conclusion, I provide below a self-designed infographic of a potential Hippocratic Oath for AI Practitioners that I felt was very apt. This “oath” was adapted from several articles from AI Magazine, TechCrunch and Yale Law School, that I found online that I thought made good sense. I would encourage all AI practitioners in healthcare to consider this “oath” from their personal and professional points of view. The ethical responsibility of how AI and Gen AI evolves will lie with everyone who is involved in the process of research, design, development, implementation and evaluations of AI technologies (including healthcare professionals, AI designers and developers, IT industry, tech start-ups, etc), so that it aligns with the humanistic values, integrity and respect that all healthcare professionals abide by.