These are the problems we want to study, drawn from building software and working with devices. Each needs a small, useful test before it needs a paper.
Proposed directions. No findings published yet.
How can we connect spreadsheets, paper, and existing software without losing records or interrupting daily work?
Start with one handover between two records. Track which fields are copied, where mistakes enter, and what a connection between them would need to preserve.
Study offline work, safe retries, and recovery on the devices people already have.
Interrupt a fictional handover, retry it, and inspect the saved record. Check for missing information and repeated actions before adding more features.
Study sensors and simple electronics with cost, repair, and local component availability in mind.
Choose one measurement. Compare a simple sensor with an available reference, then record the parts, power requirements, and conditions where it fails.
Compare accuracy, cost, and the effort of checking an assistant against a simpler tool on the same task.
Give an AI tool and a simpler approach the same small task. Keep the wrong answers and the time spent checking them in the comparison.
With permission, study access boundaries, exposed information, and recovery, then report sensitive risks to the people responsible.
Use a system we own or have permission to test. Start with account permissions and recovery, document a repeatable case, and propose a fix.
Turn a useful script, a failed experiment, or a clearer explanation into something another student or developer can build on.
Ask someone outside the project to follow the setup notes. Keep track of the missing steps before calling a tool ready to share.
AI artwork & licensed footage. Visual references, not records of our experiments.
A message might move between Egyptian Arabic, Arabizi, and English in the same sentence. We want to understand where AI gets the meaning right, where it misses, and which carefully reviewed examples would help us test it.
A possible first step: a small collection of fictional everyday messages, reviewed by fluent speakers and paired with a clear task.
A fictional Egyptian Arabic sentence is labelled, given a review note, and marked for a second review. This is an illustration of a possible workflow, not published research.
If one of these problems turns up in your work, tell us where it happens.
