Chatbots, as we all know, are computer programs designed to simulate human conversations by interacting in natural language. They can seem incredibly convenient: open the browser, go to the chat, ask virtually anything. They are VERY convenient. In a single session, you can discover a new, tasty recipe, find out who won the Oscar for best picture last year, check if the email you want to send makes sense, and much more. But as Leonelli and Mussgnug argue, we should ask “Convenient…for who?” In their paper, Leonelli and Mussgnug examine various aspects of chatbots and their claimed convenience. Here, I would like to write a short reflection on their environmental costs and the fallacy of convenience.

Chatbot’s costs, in terms of resources, are a recognised issue. I believe that Prof. Krawford is painfully correct when she says that AI is an extractive industry: from water, to lithium, and even to human knowledge and intellect. Unfortunately, the trade-off between technology and its environmental impact appears systematically forgotten when discussing AI. It’s almost a paradox; it is such a well-known issue that it continues to be routinely ignored.

Then, if institutions remain silent on this issue (and they possess far more power than you and me), what should the individual do? That is precisely what a moral philosopher is for — and much more. Philosophers are the Swiss Army knife of research! Well, I am a moral philosopher, but I don’t particularly enjoy telling people what to do. Nonetheless, I feel compelled to write a simple guide on chatbot usage. Reflecting on it, I have reached the conclusion that some tasks’ environmental costs are NEVER justified. For other tasks, we should still ask, “Is it worth it?”.

My rule of thumb is: if there is already another way to automate a task, then it’s morally wrong to resort to chatbots for it, even if it’s more convenient for the user. Some examples include automatic translation services (DeepL, Google Translate, etc.) and search engines (Google Search, DuckDuckGo, etc.). It takes minimal effort to look up online tools and prefer them over chatbots, but in doing so, we save energy and resources. Not to mention, we avoid dealing with all the other ethical issues surrounding chatbots — such as: Is the chatbot trained on protected data? Will the manufacturer take my data? Is the information received accurate, or is the bot merely spouting nonsense?

I believe this might be a simple way to start organising, perhaps by identifying ethical uses for others that are not always worth the trade-off. However, I do realise that convenience often guides action more than ethical considerations, but this is a task that is well beyond me.

Thanks to Maria for the idea of this blogpost!