[Thought experiment] If alignments are objective how do we know what they represent?

[Grek]

I think people are kinda missing the point of the question here. Saint-Just isn't asking "Does Detect Good really detect whether or not a character adheres to the IRL Objective Morality?". I mean, obviously they're not, that's an absurd thing to ask. They're asking how people would figure out the criteria by which the various Detect [Alignment] spells decide how strongly to light stuff up when cast. And for that, you'd use some behavioral studies:

Study One: Are Behavior and Alignment Correlated?
For this study, we want to know whether a specific behavior (or bundle of behaviors, if we already have a hunch about what Detect [Alignment] is checking for) will alter someone's alignment. We start out with an behavioral survey where we ask about all of the different behaviors we're curious about and we subject them to a battery of Detect [Alignment] spells. Then we send everyone on their way, with instructions to report back in next year for a repeat of the survey + alignment testing. We do statistics to see which behaviors tend to correlate to alignment changes. Maybe we find out that Detect [Alignment #3] correlates to eating brussel sprouts. Maybe we find out that Detect [Alignment #2] correlates to rescuing kittens from trees. Whatever we find, that informs the sorts of tests we do in Study Two through Four.

Study Two: Do Behaviors Alter Alignment?
For this study, we want to know whether a person can change their alignment by changing their behavior. We get ourselves a fresh group of volunteers, give them all the behavioral survey + Detect [Alignment] battery again, and then divide them up into control and experimental groups. The control group is instructed to perform some behavior that we're pretty sure isn't correlated with alignment change, while the experimental group is instructed to perform one of the behaviors we're curious about. After both groups are done with their activities, we check their alignments again. If it turns out that the experimental groups asked to eat brussel sprouts do indeed detect more strongly of [Alignment #3] than they did prior to their leafy meal, that's pretty good evidence that eating brussel sprouts will indeed cause someone to detect of [Alignment #3] out in the wild. But that's not quite enough for our purposes, because we also need to do Study Three.

Study Three: Do Alignments Alter Behavior?
This one's the opposite of the above study. We don't just want to know how to alter our alignment, we want to know what the effects of alignment actually are. Maybe there's a stereotype that people who detect strongly of [Alignment #4] enjoy violin music. Fair enough, we can test that with the Atone spell. We take our volunteers, have them subjectively rate their opinion of violin music and then split them into three groups. The control group has nothing done to them. Experimental Group A is Atoned to to [Alignment #2]. Experimental Group B is Atoned to [Alignment #4]. We play them some more violin music and have them rate that. If it turns out that Experimental Group B is the only one with a notable improvement in their opinion of violin music, that probably means that it's the alignment causing it. (If it turns out that both A and B start liking violin, that suggests that it's actually the Atone spell doing it, which would be... weird, but probably not the strangest experimental outcome in a universe where Detect Vegetable Eaters is a spell.

Study Four: Is Alignment A Good Proxy For [Insert Ethics System Here]?
This one is a bit cheeky. Nobody can agree on which ethics system is correct. But we can usually agree on which people are experts on which ethics systems. For this experiment, we check everyone's alignment, then have a panel of experts on [Insert Ethics System Here] interview our test subjects and then rate their level of adherence to [Insert Ethics System Here]. Then we check everyone's alignment again, just to be sure that being declared Highly Ethical by the Mer-Pope (or whoever else we're bringing in on our panel for [Insert Ethics System Here]) isn't one of the things that affects alignment. If it turns out that there's a strong correlation between [Alignment #3] and the having the approval of the Federated Council of Giant Spiders Who Eat Puppies, that tells us something about the relationship between [Alignment #3] and Being A Giant Spider and/or Eating Puppies.

Now, none of this actually tells us whether or not we want to be [Good] or not. After all, in 3.5 the Puppy Eating Arson Spiders From Baator will detect faintly of [Good] if someone casts Protection From Evil on them, so Detect Good is clearly not a foolproof way of deciding if someone is a decent sort or not. But equally clearly, it can at least give us some information about the probable probity of people who detect as [Good] relative to those who detect as [Evil].