MS&T readers will be aware of the increasing emphasis being placed on the capture, storage, analysis, and presentation of training data, in real time and for later exploitation. This is a trend seen across safety critical training such as for civil aviation and reflects the shift towards the wider digitisation of training. Like practically all aspects of training, the pandemic is also a significant influence as remote training reduces human interaction, putting more emphasis on other forms of training assessment. Eyes are also on the gaming and technology worlds where data is captured on a massive scale and algorithms work behind the scenes to understand behaviours and personalise and monetise experiences. AI and big data raise the prospect of personalised adaptive training and reducing contact time with human instructors.
Data captured from team and collective training, whether from live exercises or simulation, can be used to tailor training, moving away from a one-size-fits-all approach. Force elements can go through the training continuum with their training tailored and adapted dependent upon the needs and capabilities of the specific force elements. With real-time feedback to the training audience, they can be empowered to take ownership of their own development. If training data is systematically collected over time, actions and behaviours that influence performance can be identified and inform the training design.
The wider digitisation of the battlespace and increasing numbers of autonomous and automated military systems, or “boots and bots”, also feeds into the trend. Simulation can be used to bring together an increasing myriad of real and simulated data streams to aid decision support. Autonomous systems can be trained in simulation, both with and without human interaction, to optimise tactics and overall system performance. Operational and training data will come together in time as military systems and personnel learn and adjust through training and then in operations. The lessons process, force development, analysis, experimentation, research, and acquisition and support will all be better supported.
Training can be personalised and adaptive, but also its assessment can be standardised across the enterprise. “Good” training will not be determined by how long it has been done that way or because a training system is expensive to procure and support. Data-driven training can provide an enhanced, level playing field for trainees, instructors, and the enterprise.
The vision of data-driven training is extremely compelling but like most transformations there are several challenges that need to be faced. Starting from the training itself, understanding the desired outcome, and perhaps why the training is required at all. What does good training look like and who will make that decision? The more training relies on algorithms and data the more training and data experts need to be involved. It will be important to ensure that algorithms do not “mis-personalise” training where they have incomplete or inappropriate trainee data.
There are also issues such as accessibility, security, and privacy. Companies may not be willing or able to release their S&T related data, for example where they see data as a competitive advantage or something to set a high price for. Security and privacy is a whole subject in itself. Do we want all our training records held for a career or lifetime?
Perhaps the most significant challenge is data interoperability. In writing about the metaverse for this issue, I was reminded of how long the military has been able to link different simulators and live systems together and achieve network interoperability. If training data is to be shared, trusted and exploited, it needs to be extracted from its training system stovepipes and contextualized and formatted in a common way, enabling it to be combined effectively and quickly with other data. This will only happen if data is valued as a strategic asset across defence and militaries work with industry to overcome the commercial hurdles and ensure that training data is interoperable across training systems. Perhaps the next frontier of interoperability for the S&T community is training data?