A (post-) modern problem

I think postmodernism has created a generation of people who are afraid to commit themselves to an answer to a question, even when the question involves empirical data to be examined.  For example, my final exam in linguistics, just given, contained the following problem:

Review. Examine the data from Tongan (Polynesia). Then answer the questions. Assume that this data set is representative. (3 points)

[tauhi]   'to take care of'       [sino]   'body' 

[sisi] 'garland'                     [totonu] 'correct'
[motu] 'island                     [pasi] 'to clap'
[mosimosi] 'to drizzle'          [fata] 'shelf'
[motomoto] 'unripe'             [movete] 'to come apart'

Based on the evidence, the sounds [t] and [s] are: ___ Contrastive?  ___ Predictable?
Defend your choice.

Now, the correct answer is obtained by noticing that [s] always occurs before [i], and that [t] occurs before other vowels.  Thus [t] and [s] are predictable. In linguistic terms, we can rite a phonological rule:

/t/  ➔ [s] / __ [i];  [t] elsewhere.

However, I didn't require them to write the rule out.  As long as they noticed the distribution of these two sounds, they got credit.  The problem comes when, despite the admonition to take the data set as representative, stuff like this shows up (my emphasis):

[t] is predictable because it is used the same way in the middle of the text and usually following or followed by by a vowel. [s] is also following/followed by a vowel which seems to always be [i].

No, it doesn't seem as though [s] is always followed by [i]; in this data set it is the case that [s] is always followed by [i], and [t] is followed by vowels other than [i].  And then there's this:

[s] is always followed by [i] in Tongan, and [i] can't follow [t]. We don't know if they are phonemes or allophones though, it could be a coincidence that they don't follow [t]'s with high front vowels, maybe [ti] is [di] and we just don't have that data, or

I should point out that we've been working these kinds of problems all semester, and all semester I've stressed that the data sets used in the problems are to be taken as representative of the languages as a whole.  I really do think that postmodernism, despite the good things it's done for us, has damaged students' ability to commit to the solution of a data-based problem.