Is Microsoft Context IQ the Answer to Autocomplete Shortcomings?
If you use Google or Microsoft text editors, you are probably familiar with their advanced language capabilities. These editors started by offering help with spelling, grammar and synonym suggestions. Then came predictive autocomplete suggestions for words and phrases. Called Smart Compose for Google and just “Show text suggestions as I type” in Microsoft Editor, the feature helps complete words as you are writing and also suggests next words or phrases while working in email or documents. More recently, Google Smart Reply and Microsoft Suggested Replies features provide automatically-generated suggestions so you can reply to common questions in emails with just one click.
Sure, completing emails and documents faster is a productivity booster. But the vendors also tout the tools’ abilities to improve the quality of your documents. If you haven’t tried this functionality yet, it's definitely worth a shot, but know you'll likely experience an uneasy feeling of having words put in your mouth when you first start. You either get used to it or turn it off. Both Google and Microsoft offer options to turn off autocomplete.
How Does Autocomplete Work?
Machine learning underpins both of these tools. It tracks previous emails and documents to learn from previous responses before suggesting likely texts, but Google and Microsoft don't offer many specifics about how their respective autocomplete technologies work. Google Editor and Smart Compose provide “personalized suggestions [that] are tailored to the way you normally write, to maintain your writing style. Only you see your own private, personalized suggestions for your account. No other users, including administrators for your organization, can see your personalized suggestions. When personalization is turned off, you see generic suggestions as you type.”
The Microsoft Editor apparently works in a similar fashion. A 2021 article states that Microsoft “text predictions use a machine learning model to make suggestions based on the text you have typed in the current document or email.” It is unclear if input from other sources is incorporated into the suggestions the model provides.
Google admits that its autocomplete algorithm uses input from external sources. As such, the Google Smart Compose web page notes that its suggestions “may not always provide factually correct information.” It continues, “as language understanding models use billions of common phrases and sentences to automatically learn about the world, they can also reflect human cognitive biases. Being aware of this is a good start, and the conversation around how to handle it is ongoing.” In other words, suggestions provided might not only be wrong, but they may be problematic for a variety of reasons, such as displaying socially unacceptable suggestions.
Multiple studies have demonstrated how autocomplete algorithms exhibit biases, including a 2019 study that showed how they perpetuate stereotypes, and a 2022 European working paper (pdf) that explains how autocomplete influences a person’s attitude and opinion formation.
While most of these studies focus on autocomplete for web search, several recent studies have exposed serious deficiencies in current autocomplete capabilities in office productivity products. A 2021 study by Microsoft researchers evaluating autosuggestions for email replies found that “systems that ignore social context have the potential to turn otherwise appropriate replies into inappropriate ones.” A companion paper spelled out the danger: a “user might be nudged to send a suggestion to one or more users who could perceive it as problematic, potentially leading to miscommunication and damaging many relationships.” According to these (and other) papers, when autocomplete tools don’t take into account the appropriate situational context, trouble is just around the corner. Which is why I was intrigued to read Microsoft’s recent announcement about Context IQ.
Related Article: Why Enterprise AI Needs Human Intervention
Context IQ — The Next Generation of Autocomplete?
Context IQ is a new set of capabilities in the Microsoft Editor that incorporates contextual input from the Microsoft Graph to present particularly relevant autocomplete options. Microsoft Graph is a critical part of the solution because it is the underlying technology of Microsoft 365. It captures and logs events and content created by users, and then links them together in intelligent ways. For example, when you open an Outlook calendar invite, the Microsoft Graph can present the latest documents on which meeting participants are working. By mining the interactions between people and document sharing patterns for context, Context IQ should be able to surface the most relevant autocomplete options for a given situation.
Related Article: Is Viva Topics the World's First Topic Computing Solution?
How Does It Work?
The recent Microsoft Ignite 2022 Context IQ announcement explains how it works:
The Evolution of Employee Recognition
Leveraging the power of appreciation to improve the employee experience
How to Build a More Innovative and Resilient Workplace Culture
What would happen if every member of your team came to work focused on finding solutions and creating better results?
3 Secrets to Accelerating Transformation to Improve CX + EX
Learn about force multipliers that will reduce technical debt and grow revenue while reducing costs
Why Knowledge Management Is Critical to Business Resiliency
How Organizations are Future-Proofing Business by Harnessing Company and Employee Knowledge
With Microsoft Editor using Context IQ, users receive in-context suggestions while typing. While composing a message in Outlook on the web or writing in Word for the web, users can type the “@” key to invoke Editor leveraging Context IQ. A list of items (people, files, etc.) will surface for users depending on the context of the text in their message.
Continuing to type will search Microsoft 365 [i.e. the Microsoft Graph, DL] for relevant content. Selecting an item with the cursor or hitting “enter” will insert the item into the message. Results shown in this list will be sourced from the context of the current message, what is being typed, who is involved and other variables using intelligence. The preview of Editor using Context IQ is starting to roll out and will be generally available in the coming months.
One of the stated benefits of Context IQ is its ability to reduce the need to switch between apps when working, something that is known to cause fatigue, poor decision making, and even stress. If it works as advertised, this Context IQ benefit alone would be a boon to information workers.
The Bottom Line
By providing increasingly more relevant autocompletion options, Context IQ is a welcome addition to the workers’ productivity toolkit. Beyond suggesting words or phrases, Context IQ also presents suggestions for content references, like documents or people. As I have written before, context is the practical key to solving our burgeoning information overload crisis.
But there is still work to be done.
By providing people and sharing contexts, Context IQ is a good start. But in many cases, the most appropriate work context for autocomplete is the topic on which I am working, not the people I work with or the documents we share. For example, if my documents and emails are focused on a particular product design, the product’s name and associated vocabulary would potentially be the best context for providing smart autocomplete suggestions. Microsoft is working on this area as well, with Viva Topics, but progress there seems slow. The good news? The continued efforts to simplify our work patterns will surely bring relief … if they can keep pace with the complexity of our world. In the meantime, the Google and Microsoft
autoclicker er, autocomplete, features available to today may offer some respite.
About the Author
David is a product expert with extensive experience leading information-intensive technology organizations. His specialty is helping organizations “do it right the first time”— get to market quickly and successfully through a structured process of working closely with design partners from day one.