Handling sensitive topics with above mentioned example is a two-way debate gymnastics and AI still has a long way to go. Even if many chatbots are being deployed to provide help on mental problems (as mentioned in a 2022 MIT Technology Review article), the AI platforms still struggle when the topic turns highly sensitive—depression, trauma or grief, for example. They make use of natural language processing (NLP) and machine learning to identify keywords and patterns but might be short on emotional data, something a human counselor can provide. In 2020, a National Institutes of Health study determined that while AI-based tools could provide rudimentary support, they fell short when it came to more nuanced dialogue about sensitive topics with emotional content.
Platforms such as Woebot, which is an AI mental health chatbot that helps users to navigate their emotions and stressors has made its way onto the market. Woebot also has established roots in the, therefore applying cognitive behavioral therapy (CBT) principles and some studies show it can reduce anxiety by as much as 30%. Yet, some of those deeper moral topics are constrained by the limits inside its coding. However, as with all AI tools, Woebot is able to deliver genuine-sounding responses based largely on scripting; the machine has logic and text at its disposal it can use but not empathy or context — which may leave many users brushing aside deeper emotions or crises.
The sensitivity of topics that deal with personal private information makes it even more important how AI systems process such information. While healthcare AI may mismanage private patient information — especially if the system does not understand the seriousness of a sensitive health conversation, according to a report released by the European Union Agency for Cybersecurity (ENISA) in 2023. For example, a popular mental health app had its chatbot service improperly process sensitive user data, putting at risk the confidentiality of information exchanged in sessions.
AI could make services more accessible, including for mental health, but it is not yet capable of meaningfully holding sensitive conversations with emotional understanding or confidentiality that only your GP (and in some cases a therapist) are historically able to provide. An example of the same is a report published in 2021 by the American Psychological Association that reported although AI could be useful to detect early warnings for distress but it cannot replace human empathy which is an essential aspect when facing sensitive matters.
So, does that mean AI can take care of sensitive issues? Of course, AI can answer the simple questions and provide some very baseline support. But its responses might not be nuanced or sophisticated enough to handle the deeper, more complex and emotional issues. Masked AI transparency is particularly important when users engage with talk to ai, because they may have to navigate a sensitive or even traumatic conversation with the system.