Reducing Information Pollution
The following is an excerpt written by Yury Gubman, Knowmail’s Head of AI, from a recent IORG interview between Emanuele Terenzani (Lele), Yury, and myself about the need to use Artificial Intelligence to solve Information Overload.
The interview can be viewed below in its entirety, while the excerpt provides a few additional notes not available in the video.
Information overload affects individuals in different manners and communication channels, through work and personal email, social channels, and messaging platforms. Whether for work or personal life, we seem to be always connected in one way or another, just trying to keep up yet always have a feeling of falling behind. Along with technological innovation, communication just seems to be growing with more offering and alternatives, hence the problem isn’t going anywhere anytime soon.
For our own well-being, as well as an organization’s ability to maintain growth and employee engagement, happiness, and a wish for work-life balance, a solution is a must.
The main challenge is that each one of us is unique in its communication behavior, habits, and needs, hence a one-size fits all solution is not ideal, and this is where Artificial Intelligence can provide a personalized and capable solution which can grow and adapt to everyone.
Information overload is a fundamental problem everyone suffers from, which includes emails, social content and notifications, messaging, phone calls, social notifications and more. All this content pulling our attention and distracting us overloads our capabilities to actually handle matters; we can freeze unable to progress effectively, our productivity may decrease, our creativity is harmed, and even our work life balance is become more and more non-existent.
Working on an actual solution is a great challenge as you need to run sophisticated models and obtain a solution applicable to different persons and behaviors.
Dealing with a “small” data instead of “pool” data of many users, requires deep knowledge of statistical models and data exploration. As our domain belongs to social sciences, analysis of human behavior in the workplace is needed and an employee’s social network. Furthermore, these issues should be considered in every machine learning procedure.
As a statistician, I find the email overload problem as a wonderful opportunity to apply my professional knowledge and experience to the real world and improve it, as far as I can.
AI is a part of every solution!
With information overload, the reason is the inability of an employee to work efficiently with an ongoing growing number of emails arriving to his/her Inbox. AI can provide a truly personalized solution based on an individual personal behaviors and habits, granting inbox prioritization and smart filtering, much faster than a human. In fact, the AI can be seen as a digital and automated “secretary” that helps you manage your inbox much more effectively and efficiently.
Apart from AI / ML tools to help solve this problem, informational culture should be developed, especially at work. For example, reduce the amount of non-informative emails, as well as recipients, and more common sense best practices.
During your daily morning commute, you use a voice digital assistant and wish to listen to your new email. It may take hours to listen to all unread emails, but a personalized AI algorithm can detect specific emails which are urgent for you at your location, time, and state-of-mind, considering that you out of office now or aren’t in front of a computer. Knowing this, the service can summarize the email, keep it in context, as well as grant possible next-best-actions to help you actually get things done even while you’re driving, so you may reduce stress and get the ball rolling for the day.
AI uses advanced machine learning methods, with all are based on statistical theory and statistical predictive models.
Advances in machine learning algorithms allow prediction of a phenomena with very high accuracy; however, nobody can remove error term (white noise) during both data collection and model fit.
Considering user’s past behavior, his/her preferences, working times, email consumption patterns, social network and other relevant features, we can prioritize emails and provide smart suggestions for users with more than 90% accuracy. One should remember that recent research shows that human classification error is about 3%-4% (meaning accuracy of 96%-97%, so we are can close the gap in the near future.
For sure they don’t know exactly. AI algorithms are complex and some work as a “black box”. However, statistical theory provides tools to estimate an expected distribution of a phenomena we are dealing with, given available data and ML algorithm that we are trying to apply. In such a way, one can assess or estimate, a distribution of predictions. This knowledge can be extremely important in the R&D process and allows efficient data treatment and smart selection of relevant models.
That’s why a multi-disciplinary approach should be utilized in the industry, where domain data experts, statisticians and programmers are working together to tackle AI challenges.
Technology tools will continue to become smarter and smarter, but human communication will remain generally the same. One person wants to tell something to another; for this purpose, they may write a letter, call by phone, or message using WhatsApp. However, a number of communication channels will grow faster and faster, and this fact emphasizes a need in smart AI solutions in the communication domain.
Yes, for sure; one of AI’s ultimate goals is coaching. If somebody knows that an email he/she is working on is likely to be ignored by personal assistance tools, he/she wouldn’t waste time on it, or will change the text, approach, or more to increase its successful delivery as well as engagement of the recipient. As engagement after opening is also considered, it isn’t just opening an email which is deemed as success, rather time, responsiveness, and actual end-result/goals.
Plan your time. Don’t even open emails which are not important for you. Do not contribute to email overload by maintaining productive measure and common sense, i.e don’t send unneeded and/or uninformative emails. Try to think: will this email be important to recipient? If not, maybe it’s better not to send it at all?
And if you still feel the pain of email overload, install Knowmail and see how having an AI’s assistance may improve your work, and your work life balance.
No Recent What's New Posts
Our Resource Center is a crowd-sourced, moderated repository of links to IO resources of every kind. Please check it out, use what you need and contribute additional resources.
Our Resource Center is a
useful repository of research abstracts as well as links to a broad range of IO information
on the web. Learn More