To Bot or Not to Bot: It Depends on the Question

I was one of Quora’s earliest users. I earned Top Writer status for several years and even made some money through their Knowledge Prizes. Although I stopped contributing a few years ago when I felt the platform had lost its way, I still care deeply about technology-enabled knowledge sharing. Currently, I’m participating in an invite-only Q&A platform started by a couple of ex-Quorans — message me if you’re interested!

I’ve long pondered the role of human question-answering platforms versus search engines or chatbots. In a 2011 post titled “Quo Vadis, Quora?” I distinguished between objective and subjective questions, arguing that Quora was better suited to the latter. Even back then, Google was already quite good at answering objective questions. Today, retrieval-augmented generation (RAG) tools like ChatGPT and Perplexity are even better.

However, subjective question-answering remains uniquely human. You can ask ChatGPT to simulate an individual point of view, but even the best large language models (LLMs) struggle to faithfully capture human feelings and experiences. For instance, ChatGPT can generate responses to questions like “Describe your first date,” but only through hallucinations distilled from its training data — a blend of collective human experience rather than a true personal account.

Does that mean humans are obsolete when it comes to answering objective questions? Given my work as a consultant, I certainly hope not! The questions my clients pay me to answer often involve objective concerns. But we must acknowledge that machines are now more efficient and effective than humans at addressing many information-seeking needs.

Not too long ago, “let me Google that for you” became a way to mock people too lazy to search for themselves. Are we now entering the era of “let me GPT that for you”? Having others run simple queries on your behalf is generally a waste of time — unless those people are expert question askers (a.k.a. prompt engineers). If a machine can answer your question, you should ask the machine yourself.

However, if you believe your question requires an expert — or at least a human perspective — you should be prepared to demonstrate why. A straightforward classifier can determine whether a question is subjective and inherently personal, such as those involving emotions or lived experiences. In fact, ChatGPT already does this quite well.

For objective questions, you should first consult a machine-generated response and explain why you find it inadequate before asking a human for input. While this adds friction to the process, it discourages frivolous questions and ensures human effort is focused on problems that machines cannot efficiently solve.

To bot or not to bot? With apologies to the Bard, it depends on the question.