[Analysis] Will the Rise of Predictive Text become The Fall of Text Analytics? Exploring the Rise of Predictive Input Bias.

Apple’s QuickType Keyboard includes a feature called Predictive Input. It was launched with iOS 8, and with each new release of iOS, the AI that powers Apple’s Predictive Input has become more intuitive. Android keyboards have a similar feature, such as Samsung’s Predictive Text.

Increasingly, chatbots also feature predictive text. Here’s an example (photo).

Humans love shortcuts and predictive text is an epic time saver. With iOS 12, we can compose entire text message with the keyboard’s predictions. This could save someone between 20 and 40 keystrokes per sentence.

In this use case, predictive text is a self-fulfilling prophecy altering the editorial process. The predictions are displayed as suggestions at the top of the keyboard. We’d simply click on the AI’s suggested words that approximate our intent, rather than tapping out a long message on the keyboard, especially if we’re busy or don’t care much about the person or chatbot we are chatting with. In other words, a low sense of commitment skews us towards AI inputs.

For now, predictive input seems to only impact our short text messages. Not exactly one’s life’s corpus! But as the corporate chatbots rise up, it’s easy to imagine that a huge amount of our communications will be the result of predictive (suggestive) inputs. Everything from ordering pizza to our help desk tickets could be a text loop generated on both sides mostly by AI.

Transcripts from help desk tickets have been a major source of data for companies that use text classification to understand their customers. The quality of this important data will be radically altered as more of the communication from the user is a result of predictive input. It’s too early to say this will be the fall of text analytics, but it’s clear that sources of text which are free of predictive input bias will be increasingly valuable to text analysts and the companies for which they work.