With more than three decades of experience developing what would now be seen as AI-enabled learning aids, Stottler Henke has accumulated some interesting lessons learned about intelligent tutoring and after-action review, and intelligent simulated platforms and characters. 

Speaking to Shephard, Jim Ong, group manager for Stottler Henke Associates, noted that, from its founding in 1988, the company mission has focused on the development of AI applications... Continues below

This analysis article originally appeared in September's Decisive Edge Military Training Newsletter.

Newsletter Sponsor:

‘The meaning of AI has changed over the years,’ Ong said. ‘Therefore, so have the kinds of techniques that we use to carry out projects in areas like training, planning and scheduling, sensor data fusion, machine learning and knowledge management.’

As one early example, he described an effort directed toward USN tactical action officers (TAOs). Originally developed with Small Business Innovation Research funding ‘about 20 years ago’, the program was later enhanced as the TAO Intelligent Tutoring System (ITS) V2 and integrated with other software.

MVRsimulation’s VRSG: bringing military simulation training to life.

‘That was a system that had automated agents playing other members of the CIC [combat information centre] crew and also had automated coaching during the scenarios as well as evaluation afterwards,’ he explained. ‘In some of our training systems, which we call intelligent tutoring systems, the idea is to try to provide the benefits of individualised training and coaching in an automated way. And how that knowledge is encoded for use by the tutor varies depending on the application.’

‘In general, it's easier to build a focused AI system rather than one that tries to know a lot about a lot of subjects,’ he added. ‘So some of our training systems use an approach called case-based reasoning, where the tutor applies knowledge specific to the exercise for the scenario, combined with some general knowledge that that could be applied across several scenarios. And with the TAO, for example, the knowledge for monitoring and assessing the student was in the form of hierarchical, finite state machines.’

Ong pointed to the company’s development of a tool called SimBionic that supports running many finite state machines in parallel while also organising them hierarchically.

‘Each one of these boxes could be itself, a finite state machine,’ he offered. ‘But you can have potentially complicated behaviours.’

Equip Your Team with Instant Access to Defence News

Ong noted that recent research projects have focused on satellite operators, discussions among planners and the addition of automated assessment to simulated tactical situations.

‘In many training situations, it’s possible to develop automated assessment and coaching,’ he summarised. ‘And it is more feasible than some people think it is. Frequently, in a new project, the trainers will say that it’s not possible to automate something. But often we can. We pick an ambition level that that can be accomplished and use our bag of tricks.’

He concluded: ‘Traditional training, either textbook or classroom, is best at imparting knowledge. Simulation is a way of providing practice for applying that knowledge. And intelligent tutoring systems are a way to make sure that those people who are practicing are learning the right lessons, not learning the wrong lessons, because there might have been a minor imperfection in the simulation.’

Don't want to miss out on future Decisive Edge content? Make sure you are signed up to our email newsletters.