In a 1E9 survey of community members, a majority said they were friendly to AI chatbots like ChatGPT.
This behavior is also observed in science. In fact, it looks like people are even friendlier to today’s AI chatbots, which can express themselves in a very human and empathetic way, than they used to be to Alexa or Siri, which clearly sounded like machines.
But does it make sense to be nice to software and machines? Not necessarily. It can even be counterproductive.
As a study by the Institute of Software of the Chinese Academy of Sciences, Microsoft, Beijing Normal University and other research institutions found, not being friendly with ChatGPT – and probably other chatbots – can be quite helpful. Instead, it may be better to be impatient and “pressure” the service.
In fact, telling ChatGPT that a task is “very urgent” or “important to my career” appears to produce better results. In various tests, “stressed” AI chatbots performed measurably better, between eight and 115 percent, depending on the setting.
However, in the long run it could be disadvantageous to put AI under pressure or to communicate in a commanding tone, for example if children overhear this interaction. Researchers warn that this behavior could at some point spill over into interpersonal contact.
In any case, some AIs, for example Bing Chat or Inflection’s Pi, demand politeness from their users, otherwise they will not deliver any results at all.
Finally, the question to the piqd community: How nice are you to ChatGPT & Co?