Context-aware computing

AI models like GPT are increasingly used to fill in missing information. The catch: their answers can sound convincing and still be wrong.

This research builds a decision-support assistant that checks its information before it gives advice:

• It knows where each piece of information came from — an official record, a calculation or an AI model.
• It gives each piece a trust score and leaves out what it cannot rely on.
• When it does not have enough to go on, it says so and asks for more data instead of guessing.
• Every recommendation can be traced back to the information behind it.

How it helps

Wherever AI-generated information feeds decisions that matter, the assistant relies on what is verified, flags what is not and never guesses. That makes it suited to dual-use and critical-infrastructure work:

• Energy and utilities — combining sensor readings, maintenance records and AI-generated fault assessments to decide which part of the grid or network needs attention first.
• Telecom and connectivity — judging whether reports of interference or outages are confirmed or still unverified before operators act on them.
• Defence and security — building situational awareness from mixed sources and showing commanders how reliable each piece of the picture is.
• Emergency response — prioritising incidents when information arrives fast, incomplete and partly unconfirmed.

The method was developed and tested on study-planning data, where it gives lecturers an early overview of a group starting a course.