physical light switches toggle between closed-circuit and open-circuit, not high and low voltage levels. 'on', where current can flow, is a closed circuit; 'off' is an open circuit
in digital logic, when there is a correspondence between closed/open and high/low, the correspondence is virtually always that closed is low and high is open. for example: ttl inputs always treat open-circuit as high; the can bus and i²c bus "recessive" states (when nobody is transmitting) are high; avrs have optional pullup resistors on their gpios but no optional pulldowns, so if you want to connect a pushbutton or toggle switch to a gpio, you have to connect it between the pin and ground, not between the pin and vcc; 8051s' 'quasi-bidirectional' i/o ports similarly feature a weak pullup and a strong pulldown, so that, again, an open circuit is a logic high level
(and normally high is 1, low is 0, though sometimes this convention is also violated)
the only exception i know of is that 60-milliamp and 20-milliamp digital current loop interfaces, as used on teletypes, treat a closed circuit (current flowing) as a logic 1 ('mark'), and an open circuit (no current flowing) as a logic 0 ('space')
incidentally, 'closed circuit' and 'open circuit' are also confusing. a closed circuit is 'closed' in the sense that a plane curve is 'closed' if it divides the plane into an inside and an outside region. on a closed curve, you can walk around the whole circuit and return to your starting point without turning around, which is in some sense what the electrical current is doing. but of course actual electrical circuits exist in three-dimensional space, where a 'closed' curve does not divide space into an inside and an outside region, because you need a surface for that. and in all other contexts, something that is 'closed' is something that does not permit flow: a closed barn door, a closed window, a closed valve, etc. nevertheless, it is far too late to eliminate this two-dimensional flatland thinking from our vocabulary
