Randomness is introduced with each obfuscation. If the output does not look credible or the
resulting Google search is partially successful, then click the button or press Enter again.
Obfuscation adds friction to
cheating in online tests because time is a factor. It adds little friction to assignments.
Although the obfuscation works to
fool Google searches, it is useless with some other search
engines:
one click on
"Do you mean...?" and text is resolved.
When obfuscated code is plugged into MS Copilot (ChatGPT-4),
it is processed as if it was plain text. AI may have
rendered obfuscation obsolete. AI detectors like GPTzero
may now have limited usefulness with the rise of services
that sanitize AI output. Web search
q=humanize+ai+text (a popup advertisement for Seneca
came up while viewing a page at
humanizeaitext.org ) We will always
lose the arms race.
The only solution is to ensure the provenance of student
submissions, not after the fact
detection. e.g. GPTZero