Making implicit knowledge explicit: lessons from the MIT Glass Lab

Through its Education Innovation Grants, the Jameel World Education Lab (J-WEL) at MIT Open Learning aspires to develop the building blocks, ideas, and connections that power global transformation in learning. J-WEL grants support educational innovations across a rich variety of fields including: linguistics, mechanical engineering, literature, architecture, physics, management, political science, and more. More than $5 million in funding has been awarded to MIT researchers since 2017.

As part of an ongoing series, we are taking a closer look at each 2023 grantee’s projects. In the spotlight today is Andrés Felipe Salazar Gómez, a research scientist in MIT Open Learning, working closely with Alexandre Armengol Urpi, a postdoctoral associate in the department of Mechanical Engineering. Their project, “Making implicit knowledge explicit: tacit knowledge transfer from expert glassblowing instructors to less-experienced students at the MIT Glass Lab,” aims to find ways to capture expert knowledge through the use of scene point-of-view videos, eye tracking, and detailed instructions.

The focus on implicit knowledge — the intuitive know-how an expert accumulates through years of practice and training — is often difficult to simply verbalize, codify, or explicitly transfer to others. This challenge is especially evident in settings with few instructors and many learners. This project intends to create instructional resources for learners that build upon the knowledge gleaned from experts.

What excites you most about your project?

Andrés: I am truly excited about the possibilities our approach could have on the future first-time glassblower students’ learning experience. We are seeking to extract that ‘je nais se quoi’ that most experts cannot verbally explain to others, and that is only transferred through extensive training and mentorship. By instrumentalizing the glassblowing experts and their tools, we seek to quantify the different actions required for a glassblower to complete a task, and use this information to create instructional material that will facilitate comprehension of the glassblowing activities.

Alex: I am excited that we are deploying sensors in a glassblowing hot shop to monitor the key variables in play during this practice; to our knowledge this hasn’t been done before. Glassblowing is a traditional discipline with techniques that date back to Roman times, and we hope to upgrade the way it is taught and learned by adapting it to today’s technology-enhanced world.

What problem or challenge is your project trying to solve?

Andrés: As a glassblowing student myself, I have struggled for a long time with specific skills that my instructors so effortlessly master. This triggered the idea of exploring the means to capture and quantify glassblowers expertise and offer alternative resources to support the beginners learning process.

Alex: Glassblowing is well known for its crucial expert/apprentice relation and its slow learning rate due in part to the difficulties in verbal transfer of skills. There is a lot of tacit knowledge involved in this discipline, and experts usually struggle to transfer their expertise to beginners. We hope that with our project we can help structure and formalize their knowledge, and therefore facilitate and accelerate learning.

In what ways do you anticipate your project will impact its intended audience or community?

Andrés: Our approach is two-fold. We will:

1. Create instructional videos that will include the most important variables that a novice glassblower should consider when learning this craft. These resources (videos) work as support material since they reduce the cognitive load required to understand the tasks.
2. Quantify and create performance metrics of both beginners and expert glassblowers, which opens avenues to real-time analytics, feedback, and simulating and modeling the craft.

In what ways do you anticipate your project will impact its intended audience or community?

Alex: One of the outputs of this project will be instructional videos that integrate and display all the sensor information gathered in the hot shop during a demonstration of an expert. The sensed data will be shown in a visually-friendly way that is easy for anyone to interpret. We hope that this material will be used in the future as supportive training material for novices.

What does success look like for your project, and what milestones are you aiming to achieve?

Alex: There are several potential successful outcomes for our project. Firstly, we hope to find that beginners who have access to our instructional videos show a faster learning rate than those who do not. This would imply that our framework is effective in helping them structure and formalize experts’ knowledge. We would also like to discover if the use of AI-based models help us find fundamental differences in the data acquired between experts and novices when doing certain tasks. These differences could be in the way glassblowers handle and maneuver tools, the forces they apply to the glass piece, or the features they attend to. Finally, another satisfying outcome for this project would be that the instructional videos that we produce are used in the future in the MIT Glass Lab course as a class resource. This would suggest that they have a real value for students and can have a positive impact on their glassblowing learning process.

What role does collaboration play in the development and implementation of your project?

Alex: Collaboration is an important part of this project because we need to work closely with glassblowing artists and instructors in the MIT Glass Lab. This collaboration helps us understand what the key variables are that should be monitored and how to present them to novices.

What unexpected hurdles have you encountered, and how have you overcome them?

Alex: In order to not disturb the normal practice of glassblowers in the lab, we were required to use wireless sensors and be as non-invasive as possible. Moreover, we had to bear in mind that we were going to record data in a hot shop with hazardous furnaces and flaming hot glass and tools. These were constraints that made the sensor choice, their deployment and the data acquisition slightly more complex than would have been under different conditions.