Can a chatbot save lives?

Even if they can’t save lives just yet: will ChatGPT and similar tools soon become new colleagues for social workers?

Yes, I am convinced of that, although it is likely to be some time yet before generative AI – particularly in the form of text generators – is introduced into text-based online counselling. Social work still lags far behind other sectors when it comes to digitalisation, where these tools are already used as a matter of course and on a daily basis. However, it is reasonable to assume that social organisations are also addressing this issue. As counselling always involves highly sensitive topics and client data, the first step is a process of addressing organisational, professional and data protection issues.

What uncertainties remain regarding the use of generative AI tools such as ChatGPT in social counselling?

First and foremost, it must be made clear that ChatGPT itself is not suitable for direct use in a counselling context. This is primarily precluded by data protection regulations and the duty of confidentiality incumbent upon social work professionals. As all data entered into ChatGPT is stored on OpenAI’s servers, professionals would undoubtedly be breaching their duty to ensure the confidentiality of conversations.

As ChatGPT has been trained using data from the internet, a bias – that is, a systematic distortion or prejudice – arises, which is reflected in the data returned as results to a query. This can lead to discrimination and disadvantage. One possible solution would be for the social work sector to develop its own system and for professionals who are aware of data bias to train the system using appropriately diverse data. Subject to strict data protection requirements, this could result in the creation of a dedicated text generator that could also be used for counselling purposes.

What opportunities do these tools offer to those seeking advice, such as staff members?

A chatbot is available 24/7 and can be used anonymously. This is the biggest advantage for those seeking advice. It means they can get initial answers to pressing questions at any time and do not have to wait for an appointment or a response from a professional. Similarly, an AI counselling tool could be used as an everyday coach that can be accessed at any time. However, these applications differ from counselling sessions, in which a professional not only contributes their specialist knowledge on a topic but, above all, is able to perceive nuances and structure a counselling process methodically. At the same time, however, counselling professionals can certainly benefit from AI text generators, for example when writer’s block occurs during online counselling, they wish to have a text translated into plain language, or they would like to receive suggestions for next steps. In all application scenarios, however, the human remains the final authority: the professional will decide which ideas, content or text fragments to adopt and which not. And we should not forget that new mechanisms of exclusion are also emerging: not all those seeking advice have the skills and resources to use such tools.

What role will human-to-human communication play when people seeking advice often have to deal with machines in future?

I think the personality of the individual counsellor will become increasingly important. Chatbots can generate surprisingly good, empathetic and helpful responses – sometimes even better than humans. It will therefore be crucial for humans to offer added value that machines cannot provide. The term ‘emotional intelligence’ crops up time and again in this context. However, we should not overlook the fact that developments in the field of EAI (Emotional Artificial Intelligence) are already well advanced. The question “What defines a human being?” will therefore need to be discussed time and again in social work in the future. It is therefore no longer a question of “whether” AI should be used in social work, but simply of “how”.