Google recently announced in its’ developer conference new steps for their Artificial Intelligence (AI) solutions. One of them was their commitment to helping people with their virtual assistant. Sundar Pichai stated that many small businesses still don’t have capabilities for online bookings and their system will come to help them. See video of the keynote, where they presented example calls made by their virtual assistant.

Great! You can now book a haircut or a table in a restaurant with a phone assistant, that does the calling and talking for you. In short, Google has developed a system, which can fluently answer upcoming questions, take into consideration which dates are right for you and convey the pain of making the reservation. Although not quite HAL2000 or general artificial intelligence, we might be getting closer to the point in which a discussion system passes the Turing test. That is, computers being indistinguishable from humans as discussion partners.

This virtual assistant is a delightful example of how Silicon Valley pushes the boundaries of technologies. At the same time, it describes what the engineers in Silicon Valley see as the prominent problems to be solved. Your time is valuable, so better not to waste it for waiting on the phone and making a reservation at your favorite restaurant. Impressive, but I think there might be more critical obstacles to tackle, such as those in education.

 

Overloading the cognitive side with self-regulated learning

Let’s take self-regulation in learning as an example.

Self-regulated learning that is taking initiative and ownership of learning is a principal ideal to aim at. We should be able to learn and adapt to new situations, new possibilities and new developments quickly and effectively. Technological progress is changing everything at a dizzying pace. People need to learn to keep up.

Learning new things, especially when taking the initiative into your own hands, is hard. Really hard. The recently published review by Tina Seufert emphasized the cognitive load imposed for self-regulation. That is, when we take control of a new task, our minds get overloaded with magnitudes of stuff to take into consideration. Especially when you are a novice. Once we learn, the cognitive load decreases. Consider the following example.

 

Taking the driving test – from overload to automation

After receiving a qualification as a teacher and a Ph.D. in educational psychology, I finally felt confident enough to pursue a driving license. Surely, I was educated enough to tackle this one quickly. I remember my first driving lesson vividly. An empty parking lot was a frightening obstacle. The Toyota I was to master, felt as complex as a space shuttle. The order of the pedals vanishing from my mind and trying to operate the gears the same time as turning the wheel was way more than I could handle. The teacher acted as a guide. She gave me detailed instructions on which pedal to hit and which gear to choose.

Bit by bit, the pieces involved in driving got automated, and I was able to drive without constant mentoring. I passed the driving test on the first try, proof of having learned something during the lessons. Now, a little over a year later, driving seems easy, and changing gears is entirely automatic. That is, I am now a competent driver and not a cause of worry for my fellow drivers.

What happened during the process was that my skills to regulate my actions increased as some of the skills needed in driving got automated. The task of driving went from painful to comfortable. At the same time, the load imposed for my cognition decreased. The level of conscious regulation decreased from high to low. This is described in the graph below as a relation between the level of regulation and task difficulty.

 

Task difficulty and regulation (adapted from Seufert, 2018)

As seen in the above image, an easy task requires only a low level of regulation (the purple line). When the task is a bit more challenging, in the mid-level, the needed regulation increases. However, what happens when a task gets difficult is that regulation cannot keep up. As a result, the regulation level decreases. Imagine me, a first-time driver behind the wheel. Although, in theory, I knew what I should do, my mental capacity was overloaded. Fortunately, the driving instructor was a competent tutor and gave me timely instructions when needed. A tutor can act as an aid to step-up the decreased regulation. Unfortunately, a human tutor is not always available.

 

How to support the regulated learning with AI

Digital learning tools can help in two different ways: The first is learning analytics, which can portray the problematic areas for teachers, which they can focus their energy on them. The second, more futuristic way are real-time, predictive suggestions. Many designated learning applications do this with hard-coded instructions on what support to offer in which situation. But with machine learning, it is possible to generate a more general learning assistant. Three things are needed for offering such support:

  1. Understanding of the context – Technologies, such as natural language processing, image labelling and user history data give machine learning algorithms give machine learning an understanding of the setting a student is in. This is a service Google announced in their keynote as well.
  2. Difficulty level – The second aspect, the learner’s experience of challenge can be obtained with assessments, assignments and self-reports. What is noteworthy is that the evaluations do not have to be constructed by a human, but can be done by a machine as well.
  3. What supports to offer. Material suggestions from library banks. Using information from past learners’ paths to provide recommendations based on what they have done.

Taking in these pieces, we are moving towards repairing the lowered levels of regulations caused by extreme challenges with smart assistants. The future looks very interesting and virtual, indeed.

 

Task difficulty, regulation and assisted regulation

As described in the image above a tutor, virtual or physical can aid and increase the level of regulation, when needed. That is, an assisted learner can outperform the levels of self-regulation. Suggestions, explanations, detailed instructions gained from insight and experience with other learners

We at Claned have committed our AI solution to these aspects and worked hard to support the learner’s learning progress – whether self-regulated learning or supported by teacher

Topi Litmanen

Dr Topi Litmanen works as a Chief Educational Scientist in Claned Group. He is responsible for ensuring, that the pedagogical aspects of the Claned are based on latest learning research. Topi makes sure that Claned customers get the needed support for meeting their digital learning needs.

More blog posts

Cookies help us improve your user experience. To continue to do so, we ask you to allow the use of cookies.

More information about Cookies and Personal Data