The Promise of Computer-Based Instruction

I’ve been dreaming about the true promise of computer-based instruction since my days as a program coordinator for adult literacy services. The promise is adaptive and personalized learning. Two decades later, that promise still resides in the future – and in my imagination.

Imagine a responsive system that negotiates the route between your present-state knowledge and skills and your (or your school’s or your job’s…) learning goals. Such a system requires a profile of your current knowledge and skills, a map of the declarative and procedural knowledge and skills that fulfill the learning goals, and a logic engine to reconcile the two.

Why aren’t we there yet?

Several wicked problems must coalesce:

  1. Artificial intelligence and machine learning, and its human language enabler, semantic web, are nascent technologies. Sure, IBM’s Watson computer is the Jeopardy champion; but, there’s a lot more logic required for an adaptive learning application.
  2. Which leads to the problem of learning maps. We need canonical knowledge structures that trace skill enablers and knowledge dependencies. While we’re seeing innovation in item adaptive testing, currently these assessments address specific competencies categorized to grade-level standards.
  3. Plus, that level of programming logic and knowledge dependencies mapping requires a lot of time and money. Testing companies spend millions annually on item validation and reliability studies. It’s truly mind-boggling to consider the requirements for creating knowledge domain representations. No one institution could possibly undertake such a mission. We will need many committed individuals and organizations working together.

Is that all?

Of course, while enumerating the issues of computer logic, learning maps, and the necessary forces to create such a system, an essential instructional ingredient is left out. Context. A brilliant content generator without full context awareness could be, well, a Jeopardy champion (Watson’s representational and inferential efficiency is impressive, but limited to static facts). But I’m not talking about trivia facts, or about an instructional panacea. What I’m imagining are fantastic diagnostic and prescriptive tools for students, teachers, trainers and facilitators, and instructional designers. The human element must be in the mix, no matter the machine capability.

I’m still dreaming about the true promise of computer-based instruction.