Grammarly on why tone is integral to your business emails

‘AI ghostwriter’, or helpful nudge in the right direction?
25 September 2019

Grammarly’s Tone Detector. Source: Grammarly

Email, as convenient as it can be, is a tough gig.

Too direct and straightforward? That might comes across as cold, not confident. Saying ‘I think’ a lot? You might be projecting self-doubt. Let’s not get started on ‘Regards’.

By the end of this year, the Radicati Group estimates that the number of emails sent and received every day will exceed 293 billion.

Over half the world’s population now uses email, and that figure will continue to climb driven by ever-increasing internet and smartphone penetration, the rise of remote working, and an increasingly globalized workforce.

But in every email we send, we forego the facial cues, vocal intonations and tells so important to human interaction; body language and speech patterns that help us to interpret the context, or tone, behind the statement, are entirely absent.

As such, a simple line of text sent in an email— as explicit or verbose as it may be— may not always convey the intended tone of the sender. Its ultimate meaning may be misinterpreted, and the intended outcome may fly a long way past its target.

“[…] in written communication, you can’t use body language or vocal inflection, so the intent of your written message can easily be misinterpreted if you’re not conscientious about your tone,” Senka Hadzimuratovic, Head of Communications at Grammarly, told TechHQ.

The San Francisco-based Grammarly is a tech company that has developed a digital writing ‘assistant’ using AI and natural language processing.

Tone detector

Available as a Chrome extension, which works in text fields on websites like Gmail, Google Docs, and Yahoo, the ‘writing assistant’ flags spelling and grammar errors at its basic level. Its Premium version can recommend extended vocabulary, readability, and genre-specific writing styles.

But Grammarly has now turned its technology to ‘tone’, which it determines using machine learning models that look for a combination of signals, such as word choice, punctuation usage, capitalization choices, negations, and amplification words like “very” or “extremely.”

How Tone Detector works. Source: Grammarly

By analyzing these signals— word choice, phrasing, punctuation, capitalization— Grammarly can tell you how your message is likely to sound to someone reading it, offering the sender feedback before they hit ‘send’.

“In today’s workplace, we’re communicating constantly via so many written channels—email, text, Google Docs, messaging platforms—and there’s more pressure than ever to deliver a quality message in a short timeframe,” said Hadzimuratovic.

“The wrong tone can cost relationships and harm collaborative efforts. One of the hardest parts of striking the right tone is that it changes depending on the context and audience.”

Therefore, while a joyful and excited tone may be suitable when congratulating a colleague on a new promotion, it probably won’t be the best approach if asking your manager for a salary increase.

Artificial interactions?

But while the aim of the tool is to help convey the true intentions behind a message, when suggestions are generated by AI combing and processing billions of words worth of amassed text, there is an obvious risk that communications could become artificial in themselves.

Hadzimuratovic maintains that Grammarly’s tone detector only provides suggestions that users can accept or reject, instead, encouraging them to become “more thoughtful” in their editing. The company is not providing an “AI ghostwriter”, she said.

“It’s our hope that this new tone detector feature actually helps users’ writing become even more authentic. We build what we build with the goal of helping people say what they mean—which can be difficult to do without support.”

Hadzimuratovic continued: “If you receive feedback from Grammarly that your message is reading as “neutral” when your intent is to be passionate or persuasive, you can adjust the message to better engage your audience.

“If your important business message is coming off as “informal,” you can make changes until it reads “formal” or “confident.”

Of course, while Grammarly plays it down as an assistant, the development of its tone detector— and that of AI communications tools at large— could start to reduce the competitive differentiation enjoyed by naturally-skilled or trained communicators.

The impact of this technology could influence hiring choices, with less roundly-qualified candidates taking roles based on the integrity of their ‘assisted’  written communication.

The beta version of Grammarly’s tone detector is currently available as part of Grammarly’s browser extension for Chrome and will be rolled out on Safari and Firefox over the coming weeks. Users can activate by writing at least 120 characters.