Part 1 – Kalota’s Primer
Before reading Kalota’s article, I had constantly thought about why AI had to be incorporated into so many things and whether it actually worked properly. I knew it had been used to answer simple questions and generate images of people in meme form. This opened my eyes to the possibilities of what AI could do on such a small scale and at such a rapid pace, especially with ChatGPT. Once ChatGPT became popular, I knew it would eventually create new issues regarding the way AI was being used. I tried using it to generate a professional version of a picture I already had, and I quickly realized that it changed my appearance a little. I would probably be the only person to notice because I know my own face, but someone else who didn’t know me could easily believe it was real. That made me start thinking about how AI can use a person’s likeness and how these concepts are becoming intertwined within the “Terms and Conditions” that most of us automatically accept before using different platforms.
Reading Kalota’s article helped me understand what is actually happening behind the scenes. Before this, I assumed AI was searching for the correct answer the same way a search engine would, or it having the ability to alter true images and not distort them. I didn’t realize that generative AI creates responses by recognizing patterns from enormous amounts of data instead of actually understanding what it is saying, and it’s also not creating a true image, as it’s trying to copy the image but can’t properly create humans. That helped me understand why AI can produce realistic images, write essays, and create artwork that appears creative while still making mistakes that sound believable. To me, AI doesn’t truly create the way people do. It can imitate creativity by learning patterns from existing work, but it cannot replicate the emotions, lived experiences, or personal perspectives that make human creativity unique. That was probably the biggest takeaway for me because it showed me that AI is much more complex than I originally thought, but it also has limitations that people often overlook.
If a client or colleague asked me why AI “hallucinates,” I would explain that AI isn’t checking whether something is actually true. Instead, it predicts what information is most likely to come next based on the patterns it learned during training. Sometimes those predictions are accurate, and sometimes they are not. That is why AI can confidently provide incorrect information. Learning how AI works didn’t necessarily make me trust it more. It actually made me more cautious because I now understand that technology should support human thinking rather than replace it. I believe AI can be a useful tool, but I don’t think people should depend on it without verifying the information themselves.
Part 2 – Managing, Not Just Rejecting AI in Mental-Health-Adjacent Practice
I still have mixed feelings about AI in social work. I don’t think it’s something we can completely avoid because it is already becoming part of healthcare and social service organizations. At the same time, I don’t think every problem needs an AI solution simply because the technology exists. Reading Moore et al. reinforced those concerns because their research showed that AI can reinforce stigma, encourage harmful beliefs, and respond in unsafe ways during sensitive mental health conversations. Knowing that makes it difficult for me to imagine AI replacing the human side of social work. Expressing stigma and inappropriate responses prevents LLMs from safely replacing mental health providers. If an agency decided to use AI in a mental-health-adjacent role, I would insist on strong human oversight before feeling comfortable with it. I could see AI helping with lower-risk tasks like appointment reminders, client intake, answering general questions, providing psychoeducation, or helping connect people with community resources. Those are areas where AI could improve efficiency without replacing professional judgment. However, I don’t believe AI should ever make decisions about someone’s care on its own or be responsible for responding to someone experiencing a mental health crisis.
For me, the line is drawn where human connection becomes essential. Social work requires empathy, creativity, cultural humility, and the ability to understand each person’s unique story. Those qualities come from lived experience, relationships, and professional judgment, not from recognizing patterns in data. AI may be able to imitate empathy or generate responses that sound compassionate, but it cannot genuinely understand what someone is feeling or creatively adapt to every situation the way a social worker can. Moore et al.‘s findings confirmed that limitation because the models sometimes responded in stigmatizing or unsafe ways instead of recognizing the complexity of a person’s situation. Expressing stigma and inappropriate responses prevents LLMs from safely replacing mental health providers. My experiences in healthcare and clinical research have also shaped how I think about AI. I’ve learned that even small mistakes in documentation or patient information can have serious consequences if no one catches them. That is why there is always human oversight in research and healthcare. I think AI should be treated the same way. It can improve efficiency and reduce some of the administrative workload, but there should always be a trained professional reviewing the information and making the final decisions. To me, AI belongs beside social workers, but not in place of them. Its role should be to support the work we do, while the creativity, empathy, critical thinking, and human connection remain the responsibility of the people serving clients.
References
Kalota, F. (2024). A Primer on Generative Artificial Intelligence. Education Sciences, 14(2), 172. https://doi.org/10.3390/educsci14020172
Moore, J., Grabb, D., Agnew, W., Klyman, K., Chancellor, S., Ong, D. C., & Haber, N. (2025). Expressing stigma and inappropriate responses prevent LLMs from safely replacing mental health providers. In Proceedings of the 2025 ACM Conference on Fairness, Accountability, and Transparency (FAccT). https://doi.org/10.1145/3715275.3732039
10/10
I really enjoyed reading your post. You did a great job explaining how your understanding of AI changed after reading Kalota’s article, especially your example about using ChatGPT to create a professional photo. That made your point about AI changing small details and why we need to be cautious feel very real. I also liked how you explained that AI can imitate creativity but can’t replace the emotions, lived experiences, and empathy that people bring to their work.
Your discussion of Moore et al. was just as convincing. I agree that AI can be helpful with tasks like appointment reminders, client intake, and connecting people to resources, but it shouldn’t be making decisions about someone’s care or responding to a mental health crisis on its own. Your point that AI belongs beside social workers not in place of them really summed up your argument well. Overall, I thought your post was balanced, thoughtful, and made a very convincing case.
10/10
I gave your post a 10/10 because I thought you answered each of the prompts in a thoughtful way while connecting it with the text. I especially liked your point that the line should be drawn where human connection becomes vital. It made me think of another challenge, even if AI is only being used for lower risk tasks like appointment reminders or intake, those interactions can still shape a client’s first impression of an agency hence impacting their likelihood to continue seeking professional support.