There's a reason why Neurowear made the ears the way that they did: they were working with the limitations and capabilities of what their sensors could detect. The biofeedback sensors detect the wearer's inner activity along a scale of more active or more relaxed. "Interested" and "angry" are opposites in real cat ear body language, but to the biofeedback sensors, "interested" and "angry" both register as "most active."

I propose that in order to distinguish between "interested" and "angry," a hacker could add an additional pair of sensors on the wearer's brow in order to detect eyebrow movements. Raised and lowered eyebrows happen to correspond to perked and flattened cat ears. The pair of sensors would detect whether a crease in the brow has brought the pair of sensors closer to one another. The problem with that plan is that it means gluing stuff on your face.