Part one:

In my personal life, I have encountered generative AI through everyday technology without fully understanding how it works. For example, I have used voice assistants, interacted with customer service chat features, and received personalized recommendations on social media and streaming platforms without realizing that artificial intelligence was being used behind the scenes. I often viewed these features as normal parts of technology rather than systems that analyze information, recognize patterns, and generate responses or suggestions. After reading Kalota’s (2024) article, I have a better understanding that generative AI uses technologies such as natural language processing and large language models to create content and respond to users. This has helped me become more aware of how AI is already present in my daily life and the importance of understanding how these tools work, as well as using them responsibly.

Understanding how generative AI works makes me more comfortable with the idea of AI becoming part of social work practice, but it also makes me more aware of the need for responsible use. Learning that generative AI uses technologies such as large language models to analyze patterns and generate responses helped me understand how these tools can support social workers with tasks such as organizing information, improving efficiency, and locating resources. However, Kalota (2024) also explains that generative AI has challenges related to accuracy, privacy, ethics, and the possibility of relying on technology too heavily. These concerns are especially important in social work because practitioners work with sensitive client information and support people experiencing complex situations. I believe AI can be a useful tool when used ethically, but it should not replace the empathy, professional judgement, and relationships that are essential to social work practice. Understanding how AI functions allows me to see it as a resource that can assist social workers while keeping human connection at the center of the profession.

If a client or colleague asked me why an AI tool sometimes “hallucinates” or provides information that is confidently wrong, I would explain that AI does not understand information the same way people do. I would say that AI is designed to recognize patterns and create responses based on the information it has learned, but it does not actually know whether every response is true. Sometimes, it may provide an answer that sounds accurate because it is following patterns in the data, even when the information is incorrect. I would explain that the AI is not purposely trying to mislead anyone; it is simply a limitation of the technology. Kalota (2024) explains that generative AI responses depend on the data used to train these systems, which means the information it produces should always be reviewed and verified. I would remind clients and colleagues that AI can be a useful support tool. Still, it should not replace human judgement, professional expertise, or the careful attention needed when working with people’s lives and experiences.

Part two:

If I were considering using an AI tool with real clients, one safeguard I would insist on is that a trained mental health professional must always provide human oversight. I would explain that AI can be a helpful resource, but it should not be relied on to make decisions about a client’s care because it may not fully understand a person’s emotions, experiences, or individual needs. A professional should review the AI’s responses and step in when a situation requires empathy, critical thinking, or professional judgment. Moore et al. (2025) found that AI tools can sometimes provide inappropriate responses and show stigma toward people experiencing mental health concerns, which highlights the importance of having human involvement when using these tools. In social work, I believe AI should support the work of professionals, not replace the relationships, compassion, and ethical responsibility that are necessary when working with clients.

The line between an appropriate task for AI and a task that requires a human is based on whether the situation requires professional judgment, empathy, and the ability to respond safely to a client’s needs. I think AI may help provide general support or information, but it should not replace a human professional when working with clients who have complex mental health concerns. Moore et al. (2025) showed that AI tools can sometimes provide inappropriate responses and demonstrate stigma when responding to mental health-related situations, which shows the importance of human oversight. Due to these limitations, I would only feel comfortable using AI as a supportive tool with clear boundaries and continued involvement from trained professionals who can evaluate responses and make decisions about client care.

During my field placement, I have seen how important oversight and human involvement are when supporting clients and families. Social work decisions often require understanding the full situation, listening carefully, and considering factors that may not be obvious at first. When there is a lack of oversight, important details can be missed, which may lead to responses that do not fully meet a person’s needs. This has shown me that whether it involves AI or any other system, there needs to be accountability and someone who can review decisions and make sure they are appropriate. I believe technology can be helpful, but it should always be used alongside human judgment, empathy, and professional responsibility.

Reference:

Kalota, F. (2024). A primer on generative artificial intelligence. Education Sciences, 14(2), 172. https://www.mdpi.com/2227-7102/14/2/172

Moore, J., Grabb, D., Agnew, W., Klyman, K., Chancellor, S., Ong, D.C., & Haber, N. (2025). Expressing stigma and inappropriate responses prevents LLMS from safely replacing mental health providers. In Proceedings of the 2025 ACM Conference on Fairness, Accountability, and Transparency (FAccT’25) (pp.599-627). Association for Computing Machinery. https://dl.acm.org/doi/epdf/10.1145/3715275.3732039