The ear is not flat. At moderate levels it hears the midrange, roughly where voices live, far more sensitively than deep bass or the highest treble. A-weighting bakes that curve into measurement: filter the signal to mimic the ear’s sensitivity, then read the level. The result is quoted as dBA, and it is what noise regulations, venue limits, and handheld SPL meters almost always mean.
The filter’s shape follows the ear’s response at fairly quiet levels, which is the standard criticism: at club volume the ear flattens out and hears bass much better than A-weighting assumes, so a bass-heavy system can hammer a room while producing modest dBA readings. C-weighting, a much flatter curve, exists partly to catch this, and serious venue measurement quotes both.
For anyone building or measuring reactive systems the lesson carries over: how loud something reads and how loud it feels are different questions, and every metering decision, dB, dBA, or LUFS, is a choice about which question to answer.