Correct behavior is not the same as acceptable behavior
A robotics engineer can solve the mechanics of a task completely, an arm that never misses its target, a mobile base that never collides with anything, a gripper that never drops its payload, and still end up with a robot that people don't want in the room. That gap exists because 'working correctly' and 'feeling right to be around' are different problems, measured by different criteria. Mechanical correctness is judged against physics: did the robot achieve the state it was commanded to reach, within tolerance, without damage. Human acceptance is judged against a much fuzzier and more subjective standard: did the person nearby understand what was happening, did they feel safe, did the robot behave the way a reasonable observer expected it to. A robot can score perfectly on the first axis and fail badly on the second.
This is why HRI exists as a field with its own methods, vocabulary, and failure modes, rather than as a footnote to controls or perception. It draws on psychology to understand how people form expectations, attribute intent, and react to ambiguity, and on design to translate those insights into concrete choices about motion, sound, appearance, and timing. A team that only staffs control engineers and roboticists will tend to optimize for the metrics they know how to measure, speed, accuracy, uptime, and treat human comfort as an afterthought to be patched in late. Treating HRI as a core discipline, present from the start of a robot's design rather than bolted on after the mechanics are settled, is what actually produces robots people are willing to work alongside.