Part 1 

Kalota’s article brought up a ton of concepts about AI I have never heard of. One AI concept that stood out to me was the distinction among AI, machine learning, and deep learning. Before reading this article, I thought of AI as a single technology. However, I now know that is far from the truth. Kalota explains that generative AI is built on different layers of technology, with large language models using deep learning and artificial neural networks to recognize patterns and generate responses rather than actually understanding the language. The explanation of how GPT predicts the next most likely word, rather than solving problems like a human, corrected my mistaken belief that AI had a process for thinking through problems like humans. I have encountered AI in a plethora of different avenues. I have personally used AI during my placement to help me develop appropriate SEL lessons for different age groups. I have also come across AI in a few of my courses. Some of my professors have openly incorporated AI in coursework as a tool to be used. For example, one professor I had allowed for the use of AI to help with assignments; however, you needed to provide a detailed summary of what AI you used and what exactly you prompted the AI to do to help you with said assignment. Without the AI doing it for you. Since I now understand a little better how generative AI works, I’m more optimistic but also more nervous about AI’s role in social work. When Kalota was discussing the potentially inaccurate outputs and black-box decisions without any explainable AI, it reinforced for me that AI should be used in a support role for social work. I also learned that AI is only as reliable as the data it is trained on. Therefore, even elaborate systems like Meta’s LLM Galactica can be proven to be riddled with false information. Therefore, in the social work profession, our decisions have a powerful effect on people’s lives; the limitations Kalota speaks about are too significant to allow AI to be more than just a support role for social workers.

Part 2 

Reading Moore et al. further reinforced my previous thinking that AI should not replace real social workers or therapists. I agree, though, that the social work profession shouldn’t reject AI completely, especially because it is already being used in our profession and its presence will only grow. The article’s research found that large language models can show stigma toward people with mental health conditions and in turn provide unsafe responses that may include things like reinforcing delusions and responding inappropriately to suicidal statements. This just can’t happen; therefore, the biggest safeguard that I would require would be to have a mandatory review by a licensed, trained social worker or therapist whenever the AI interacts with clients. AI shouldn’t make decisions regarding risk, diagnosis, treatment, and crisis intervention by itself. I think it is okay for an AI tool to be used for things like intake, psychoeducation, and afterhours support; however, every interaction that involves safety concerns should be immediately reviewed by a qualified professional. I also think that if agencies are going to utilize AI technology, they should have the resources to also regularly test the system for things like bias and accuracy before it is implemented with clients.

In my opinion, Ai can absolutely help with administrative tasks that would improve the efficiency of the professional without replacing the judgment of that professional. While things like crisis interventions, diagnosis, suicide risk assessment, etc.—basically all things that require ethical decision making and therapeutic relationships—should be performed by humans without AI. Due to Ai’s lack of emotional intelligence and ability to connect with other humans. During my school social work internships, I saw firsthand how important human oversight is when responding to students who are in crisis. these tough decisions that involve different family dynamics, safety, and compassion are what only a human can do. Therefore, AI can support social workers but can’t replace them.