A criticism has emerged that passing off writing produced using generative artificial intelligence (AI) as if it were personally written can cause others to misunderstand.

On June 12 local time, online outlet Gigazine reported that Siavash Sahebi and Thomas Montefiore, who study philosophy at Australia's Macquarie University, examined what ethical problems can arise from communication that does not disclose AI use.

The point at issue is whose thoughts, abilities and emotions the other person takes the writing to reflect. Comparing a case of summarising meeting content with a case of a eulogy, the researchers said the ethical weight of not disclosing AI use varies depending on the situation.

For example, consider a case in which someone uses AI to summarise meeting content and shares it with colleagues. In that case, others may judge the writer to be highly skilled at summarising. But if AI performed a substantial part of the work, not disclosing it can lead to misunderstanding about the writer's role and abilities. The researchers said such cases can raise ethical issues, but it is hard to say that every work process must always be disclosed.

A eulogy, by contrast, is treated as a more sensitive case. If a longtime friend reads a eulogy honouring the deceased and it is later revealed that AI wrote it, the ethical weight can be much greater. The researchers explained that this is because the audience understands not only the content of the sentences but also that the writing contains the friend's own feelings and wishes.

Behind this awareness is existing research showing that disclosing AI use affects human relationships and evaluations. Research has already been presented suggesting that when people disclose that they used AI, evaluations of the writing fall or there is a tendency to rate the abilities of the person who used AI lower.

Citing philosopher John Danaher’s “classification of deception by AI and robots,” the researchers divided AI-related deception into three types. These are a type that induces misunderstanding about the external world, a type that causes people to misperceive one's abilities or achievements, and a type that conceals one's true thoughts, emotions or abilities. The meeting-minutes case is closer to a case that causes others to misunderstand the work actually performed and the abilities demonstrated.

Even with the same tool, the scale of the problem varies by context. Spell-check functions may not pose a major issue without disclosure in general writing, but using them in a spelling test is a problem.

There are exceptions even for a eulogy. If a friend is in such deep grief that writing directly is difficult and AI help is needed to meet a funeral schedule, that should be taken into account.

Ultimately, the point at issue is whether hiding AI use has a significant effect on the other person's judgement. The researchers said that in situations where the other person must accurately understand the speaker's thoughts, emotions and abilities, it is important to disclose whether AI was used. They added that what constitutes a significant misunderstanding can vary depending on social norms and the communication context.