Challenges in understanding AI

Kjell Staffas

Erik Bjurström

Git Roxström

Understanding AI in the context of C2

• What do future officers need to understand about AI in C2 to make decisions for decisive advantage?
• Which are the crucial challenges to the education of AI-savvy decision-makers?
• Which are the greatest challenges to the actual realization of the full potential of AI-supported C2?

Risk-savvy decision-making

• Risk – decision-making situations where all potential outcomes and their probability to occur are known
• Uncertainty – situations where either the outcomes and/or their probability to occur are unknown
• For risk-savvy decision-making you need to calculate the risk as well as the uncertainty for all situations and parametres to make good decisions
• Hence, in C2 decisions, the unknown are typically as important as the knowns of a situation

”How to make good decisions”

• Rely on intuition and heuristics

- preferable for simple problems when the uncertainty are high, there are many alternatives, and the amount of data are small

OR

• Rational decision making

- Complex problems benefits from low uncertainty, few alternatives, and a high amount of data

The need for an AI literacy

• What can AI do and not, especially in relation to human intelligence?
• Modelling C2 in AI-supported systems is to understand how to evaluate information, knowledge and experiences as risks and uncertainty
• AI is suitable to calculate risk and uncertainty for further processing to aid if simple heuristics or complex models are the best suited
• In the light of The Chinese room argument; there are no such thing as awareness in AI responses. It just responds from processing an enormous amount of data.

Notations on the understanding of AI and its effects

• Understand what AI can do better
• What can human experts better than AI, and to what extent can AI support this knowledge
• How works human intelligence
• When the level of uncertainty becomes too high heuristics and intuition are at least as good as extensive analysis.
• The ability to predict adversaries´ AI power will be crucial
• AI systems as team members
• Evaluation skills become one of the greatest assets for the future C2 officer, i.e. risk-savvy decision-making

​

Future agenda for research on AI-supported C2

• Understand AI compared to human consciousness and capabilities
• The methaphorical understanding of AI vs. Human intelligence is crucial for the further use of the technology
• Highlighting the difference will provide for future research on how to increase the probability of making AI useful for C2 decision-making