After 60 years of computing power growth supported by Moore’s Law, and 60 years of AI development, we’ve entered an Industry 4.0 era, with connected networks of everything including data. The speed of innovation, competition, and change is unprecedented. The pace of iteration can be as short as hours with automated iterations by Machine Learning. (as below, image credit: Robin.ly) The question for every organization is: How fast are you iterating toward your goal?
In the midst of the COVID-19 crisis, it is more urgent than ever that modern organizations be agile and efficient in responding to a rapidly changing environment and competing to survive. A learning organization is a company that facilitates the learning of its members and continuously transforms itself. [1] Now the speed of organizational learning, which is an orchestration of an organization’s employees with structures and processes in place, needs to accelerate. Corporate L&D (learning and development) department plays an important role.
To facilitate that acceleration, iterations of data are required. In our interview with Nate Hurto, SAP, he emphasized the importance of iterations loops, macro and micro, needing to be as fast as possible, as being fundamental to an intelligent enterprise. Nate Hurto reflected that when these intelligent organizations start to “combine data from different systems, there is a transformation that happens.” [2]
Learning Engineering and Organizational L&D
In marketing technology (MarTech), highly granular personalized or persona-targeted recommendations have been changing user behavior for many years. Those techniques have often been used for political causes as well. In driving a learning organization to accelerate learning, by supporting every employee to learn faster, there is still much to be done. The macro and micro iterations in current enterprises, even in a lot of Learning Management Systems (LMS), usually aren’t built with the science of learning in mind.
Enter “Learning Engineering”.
“Learning Engineering is a process and practice that applies the learning sciences using human-centered engineering design methodologies and data-informed decision making to support learners and their development.” — IEEE ICICLE
Behavioral science interventions have been developed to promote a variety of prosocial behaviors, such as healthy eating habits, physical activity, getting medical check-ups, voting, and achievement in schools and colleges. Learning Engineering incorporates solid learning science, pedagogical best practices (based on a lot of research done in the past few decades), empirical approaches, and design thinking as well as teamwork. To learn more about cognitive principles for designing effective remote learning, check out The Science of Remote Learning by Goodell, J. & Kessler, A. (2020).
Learning Engineering combined with an algorithmic approach plus machine learning leads to true power and speed in pushing learning in an organization forward . This combination of engineering principles with algorithmic solutions and machine learning is already happening in other engineering disciplines. This combination could be called “AI for learning” for short.
Considerations for L&D in the Corporate Context
Let’s re-examine L&D in the corporate context, which is more complex than in K12 or higher education contexts.
- L&D is a means to an end and is part of a bigger system.
- Employees have no time to learn, and real learning often happens while working.
- “Saving employee’s time is the golden KPI, and what is desired is enabling learning in the workflow”[3] (Interview with Alban Jacquin, Learning Experience & Innovation Director, Schneider Electric) The key question is how to enable that.
- “L&D team needs to be able to measure the impact of L&D”, according to Alban Jacquin, his team measures and monitors 100+ metrics. [3] Being able to correlate L&D and business outcomes and know ROI of investment in R&D is desired by executives.
- “Better skill assessments are much needed”, according to Alban. [3]
- Issues about privacy, data integration, and safety can’t be neglected. Hopefully, we want to have more than data in learning applications. According to Nate, “We are a privacy conscientious company, so we don’t log granular behavior events”.[2]
- New knowledge is constantly discovered and created at work or in interactions between people, across many tools or systems. How can this be captured? (It seems Microsoft Project Cortex is aiming to solve this problem, and is providing resources for developers).
- AI has already started to make an impact in corporate L&D [4] (Interview with Avinash Chandarana, Head of MCI Institute, MCI Group),
- L&D departments need to become the performance consultant and learning experience architect for the business, not only order takers, according to Alban Jacquin [3]. Businesses need to consider what the L&D departments need to catch up, or how the business can support and augment L&D’s work with technology.
- Enterprises have been facing challenges of digital transformation as well as skill gaps for the AI/automation age, the COVID-19 is only worsening it.
AI for Learning
In an interview with Chris Littlewood, head of data science at Filtered, he explained how his startup created an “AI for Learning” solution for enterprises. Filtered needs to work closely with enterprise L&D teams to build a skill framework that’s aligned with their goals. The more specific each skill definition is, the better. L&D teams need to be able to curate and label a certain volume of content to the skill framework for training purposes. From there, Filtered can use Machine Learning to infer and make recommendations. The system interacts with users via questions to understand users, and tries to collect other kinds of data besides learning data for better modeling. Chris reflected on the lessons learned from their journey, and realized that although AI is fancy, if enterprises don’t realize that its impact links to business survival, the project will only be categorized as“innovation” (meaning not urgent). [5]
The value brought by Filtered is mainly content discovery. It’s not an intelligent tutoring system (or adaptive learning system) yet, but its impact has been much appreciated by its clients, among them the MCI Institute. Avinash Chandarana, Head of MCI Institute, mentions that their L&D team needs only two people now. The AI engine by Filtered increased the engagement of learners significantly, and enhanced the productivity of the L&D team. Avinash even uses Filtered to help MCI’s clients to engage their own clients in online events. It has been very much needed during the COVID-19 crisis, as almost all in-person events are canceled.[4] In a learning context, AI supports learner agency for self-paced learning and it’s especially needed now since remote work has become the new normal.
Alban of Schneider Electric notes that they have also used an AI engine to recommend learning activities for learners. A set of skill models for over 800 job roles is the foundation of their AI model. The AI engine inferencing can help learners more dynamically because the skill framework isn’t updated frequently. Both push and pull approaches are necessary. Learning in the workflow is a new thing for them, and they are trying to push in that direction (we can spend a whole day talking about this). Alban does hope to be able to better assess skills.[3] We see that on many CLOs’ wish lists.
What then is needed in order to build an intelligent tutoring system, for effective personalized or micro-targeted learning at scale? A good conceptual framework is to refer to the Total Learning Architecture (TLA) by ADL, which always pushes the boundary of human capability development because of its mission to train the military with efficiency. In short, there are four main pillars to Total Learning Architecture (TLA), listed below, plus pedagogical engineering. There are also quite a few relevant works being done and ongoing at IEEE society (such as IEEE AIS Standards)
- A competency framework (aligned with business goals in the corporate context)
- Knowledge modeling (for better sequencing and structuring of the learning process)
- Learner modeling (to model many dimensions of a person)
- Experience tracking across relevant systems (this is where Experience API (xAPI) plays its special role. xAPI enables us to collect behavior data at work in order to link learning and performance).
Remember, learning technologists need to look outside of learning scope, use technologies to level up organization performance, and be able to measure its impact on business outcomes.
Management Strategy is an Important Dimension
To drive the acceleration of learning for employees, there is another dimension of management strategy we need to consider. Andrew Saidy, VP Talent Digitization of Schneider Electric, implemented Open Talent Market (OTM) a year and a half ago. OTM has had a big impact on retaining talent, matching dynamic needs in the organization with available talent, driving mobility in the company and motivating learning. OTM links to an LMS and this implementation helps employees own their learning.[3] OTM is the management strategy which creates a drive for learning when people want to change or level up their career. To make OTM work, there must be a good modeling of skills for jobs and employees, so the AI engine can do match-making in less than 12 seconds. A great strategy needs the support of good engineering.
Think Bigger about What’s Possible with AI
In a survey done by Wiley, 55% of employers surveyed want to rely on AI to close skill gaps. There are several possibilities.
- “AI for learning” can help with learning/training efficiency (1st thought),
- AI can augment worker performance (there should be a good integration of learning and work, imagine an AI assistance integrating learning, performance support and actionable data ),
- AI/automation can replace human workers in completing aspects of their jobs, so where there is not enough talent, maybe let’s have AI help.
From a business perspective, if a task can be done by a machine, it could result in significant savings and extreme speed. What machines are capable of is growing every day. In certain kinds of tasks, machines perform much better than humans. All that machines need are great algorithmic modeling and problem-solving strategies. That’s why we should keep in mind the opportunity of collaboration between humans and AI.
Repetitive tasks can be done by Robotics Process Automation (RPA), but higher level tasks can be automated too. For example, there is not enough talent capable of building and deploying Machine Learning models now, so many “AI builds AI” tools, based on AutoML, are emerging. They reduce the demand for data scientists and enhance the efficiency and quality of modeling and AI deployment. Machines can carry out trial-and-error steps, workflow of best practices, and logical reasoning on data and knowledge, among others. Now AI can assist with the highly-skilled and complicated IC design process, and beat experienced IC designers in place-and-route circuit layout because there are too many possibilities and too complicated.
AI Enabled Learning Organization
How fast an organization can effectively iterate toward goals is the decisive factor for thriving in these uncertain times. Key factors for that success are (1) data & integration (2) modeling & a problem-solving approach and (3) valid measurement. We have talked a lot about the opportunity of a Learning Engineering approach to drive a learning organization, but what is also important is the collaboration between AI and humans for overall performance. There are still a lot of opportunities left at the intersection of learning engineering, AI/automation, and data-driven enterprise operations.
Invitation for Case Study Submission
As enterprises are facing the pressure of digital transformation right now, we invite you to submit your case studies for this program intended to surface best practices or solutions from problem-solvers, especially bold startups — Corporate Digital Transformation Enabled by L&D or Augmented Workers. We also like to invite enterprise leaders and investors to join our panel discussions that build knowledge together. Enterprises can put out their specific quests.
Email for participation: Contact[at]WiseOcean[dot]Tech
References:
[1] Pedler, M., Burgogyne, J. and Boydell, T. 1997. The Learning Company: A strategy for sustainable development. 2nd Ed. London; McGraw-Hill.
[2] Looking at Big Picture of Intelligent Enterprise, interview with SAP
[3] Talent Digitalization & Digital Learning, interview with Schneider Electric
[4] Digital Transformation Driven by AI Recommendation Engine, interview with MCI Institute
[5] Skill Inferencing and Autonomous Learner, interview with Filtered
Authors:
Jessie Chuang, Co-founder of Wise Ocean & Classroom Aid, Vice-chair of IEEE ICICLE
With a background in Science and Engineering, high-tech. Industry R&D and consultancy experience with corporations on learning technologies and AI, Jessie has been dedicating to build knowledge networks for big challenges and “AI for Executives programs” for AI digital transformation.
Barish Golland, Training Lead for Enterprise Resource Planning (ERP) implementation at the University of British Columbia
With a 20 year background in education, teaching, educational technology support & learning technology ecosystem management, Barish is passionate about enabling organizations to leverage the best-in-class learning technologies to empower employees with the training and readiness they need to succeed.
Jim Goodell, QIP, Vice-chair IEEE LTSC, IEEE ICICLE Steering Committee
Jim is a thought leader in the world of learning engineering and data standards, Vice-Chair of the IEEE Learning Technology Standards Council, and Senior Analyst at Quality Information Partners. He chairs the IEEE Industry Consortium on Learning Engineering Competencies, Curriculum, and Credentials SIG, the IEEE Competency Data Standards Workgroup, Adaptive Instructional Systems Interoperability Workgroup, and serves on the ICICLE Steering Committee. He leads the development of CEDS.ed.gov data standards with QIP and AEM. With the U.S. Chamber of Commerce Foundation, he co-led the development of the T3 Innovation Network’s LER Wrapper specification.
