Overestimating AI’s communication ability is bad for business 

“I give this AI-written presentation a 9 out of 10. It’s much better and clearer than my English presentations.” “I give it a 6.” “I think it is a 7.” 

We can learn something important from these three responses of Taiwanese sales specialists in my recent Taichung sales presentation skills workshop after I introduced them to ChatGPT and Bing Chat.    

For the low- to mid-ability language user, the AI text will be better than what they could manage by themselves. This is a powerful tool that can truly “democratize” international communications in English. With some training in making clear prompts, the user with a low language or communication ability can produce acceptable and very intelligible communications. 

So, what about high-intermediate and low-advanced level communicators, like many of non-native English writers in the in the sales, marketing and PR departments in Taiwan’s large international organizations?    

Although their English language ability is quite impressive, their communication ability is typically locked at a sub-professional intermediate level.   

This means they are especially susceptible to the Dunning-Kruger Effect: people of lower ability often overestimate their ability.  

And this can be exacerbated by using AI. By using and even depending on AI, these users may succumb to what we might call an AI-assisted Dunning-Kruger Effect where the language looks and sounds accurate and fluent, but the communication may contain style, tone and word choice issues that can make the writing or speech sound inappropriate or unprofessional.  

This is not always important, but for high stakes communications like press releases, speeches, and diplomatic letters, this can make the organization or brand sound unprofessional, or worse, offensive to the target audience. 

If you need to communicate in a high stakes or professional context, AI may give you a text that meets the bare minimum requirements. But perhaps less. This is a real issue when an organization’s reputation is on the line.  

For effective professional communication, you need to understand two issues when using AI to enhance your communication ability: The specifics of the context, and the ability level of the user. 

The first issue involves the context’s communicator and target audience. The level of tone, formality, and specificity of information are key considerations for an effective message.  

The second is the language and communication ability level of the user. For low and high ability users, AI chatbots can provide clear gains in ability or productivity.  

Recent research found that AI tools like ChatGPT can help a lower ability communicator become an intermediate level one for tasks like email and report writing. For a high ability user, AI chatbots can save a lot of time creating raw material and even providing interesting new ideas that the user has the ability to evaluate.  

For intermediate level users, the situation is more precarious.   

I work part-time as a proofreader in a large Taiwanese non-profit organization to make sure that speeches and diplomatic correspondence have the correct language, fluency and tone.  

Although the writers of these documents have impressive English ability, they are not trained in diplomatic or business communication and are thus intermediate level communicators. Often, they have to write high stakes communications that could affect the reputation of the non-profit. 

These writers have for several months been using ChatGPT to revise their writings. On the positive side, I have now noticed fewer linguistic mistakes; however, I have also noticed some new problems that are more broadly related to communication style and appropriacy. 

In terms of language style, AI chatbots like ChatGPT and Bing Chat prefer a generic, repetitive and concise writing style. They also tend to use similar vocabulary and sentence patterns. One recent letter I revised used the same sentence pattern (a sentence with a list of A, B, and C) in four sentences out of five. 

Two weeks ago, a writer came to me after her formal invitation letter was rejected several times by the Chairperson. The grammar was correct, and the language was fluent. But there were two important problems with the communication: It was twice as long as it should have been, and the main point of the letter was hidden in the third paragraph.  

Last week, I edited a recent thank you letter to keynote speakers and immediately noticed a paragraph with adjectives and empty praise for their speech. It was clear the writer had not seen the speech, which is not the message the Chairperson would want to convey.   

A long-time issue I have noticed in emails and speeches is a slightly arrogant tone and position if the audience is being helped by the non-profit organization. This was obviously not intended. 

These new problems were overlooked by both the AI and writer, but probably not the reader.  

AI should be an indispensable tool in a professional’s workflow—for low- to advanced-level language users and communicators.  

AI training is crucial, but it is not sufficient for high stakes communication. An expert human in the loop is still required if the quality of communication and reputation of the for- or non-profit organization is considered important. 

The first step is to realize the Dunning-Kruger Effect and guard against it. Otherwise, we are blindly following the AI systems that can score a 9 out of 10 on language accuracy but only a 6 out of 10 on communication quality.

Leave a Comment

Your email address will not be published. Required fields are marked *