I love your idea in principle, but it would be very problematic, and very sensitive to the tester's prior experience. I think that many people would (un)naturally be more trustful of mathematically-looking squiggles than a dog's barks, while others, especially if they got burned by damn lies before, would inherently distrust mathematics. In any case, people are very prone to trying to complete gaps in their understanding of a speaker based on prior conceptions.