TrustFormer: New AI Helps You Choose Trustworthy Collaborators

In any group effort, from a work project to an online gig, choosing the right partner can make or break the result. But how do you know who to trust? Most current systems use a single score or rating, which misses the full picture. Someone might be fast but sloppy, or friendly but unreliable. TrustFormer, a new AI system, looks at multiple sides of trust at once — like dependability, skill quality, and timeliness — and tracks how those change over time.

The system works by pulling together data from past collaborations, even when that data comes in at different times and in different forms. It uses a technique similar to the one behind ChatGPT, called a Transformer, to spot patterns in how a person's behavior evolves. This lets it build a detailed trust profile for each potential teammate. Then, when a new task comes in, TrustFormer compares what the task really needs — say, speed over perfection — with each person's profile to find the best match.

The results are promising. In experiments, TrustFormer beat existing methods by a solid 40.8% in trust evaluation accuracy. That means fewer surprises like missed deadlines or sloppy work. This could be useful not just for work teams, but also for online marketplaces, crowdsourcing platforms, and even autonomous systems that need to cooperate safely.

Of course, this is research, not a ready-to-use app yet. Trust evaluations are only as good as the data they're built from, and no algorithm can guarantee a person's future behavior. Still, TrustFormer points toward a smarter way to decide who gets your trust.