Comment by eviks
14 hours ago
> This is the only weakness of Barr's method.
Of course it isn't, the other weakness you just dismiss is the higher risk of failed searches. People already fail with digital, it's even harder to do in print or translate digital to print (something a machine can do with 100% precision, now you've introduced a human error)
> At the end of the day, you need people who understand the context
Before the end of the day there is also the whole day, and if you have to waste the attention of such people on doing ink redactions instead of dedicating all of their time to focused reading, you're just adding mistakes for no benefit
> something a machine can do with 100% precision
Forget about typoes. Until recent LLMs, machines couldn't detect oblique or identifying references. (And with LLMs, you still have the problem of hallucinations. To say nothing of where you're running the model.)
> if you have to waste the attention of such people on doing ink redactions instead of dedicating all of their time to focused reading
You've never read a text with a highlighter or pen?
Out of curiosity, have you worked with sensitive information that needed to be shared across security barriers?
Reading through material in context and actively removing the telling bits seems very focused to me.
Furthermore, reading through long winded, dry legalese (or the like) and then occasionally marking it up seems like an excellent way to give the brain short breaks to continue on rather than to let the mind wander in a sea of text.
I am for automating all the things but I can see pros and cons for both digital and manual approaches.
The reading is focused, but that focus is wasted on menial work, which makes it easier to miss something more important
> give the brain short breaks
Set a timer if you feel that's of any use? Why does the break have to depend on the random frequency of terms to be redacted? What if there is nothing to redact for pages, why let the mind wander?
> I am for automating
But you're arguing against it. What's the pro of manually replacing all 1746 occurrences of "Trump" instead of spending 0.01% of that time with a digital search & replace and then spending the other 1% digitally searching for variants with typos and then spending the last 99% in focused reading trying to find that you've missed "the owner of Mar-a-Lago Club" reference or something more complicated (and then also replace that variant digitally rather than hoping you'd notice it every single time you wade through walls of legalese!)
> What's the pro of manually replacing all 1746 occurrences of "Trump" instead of spending 0.01% of that time with a digital search & replace and then spending the other 1% digitally searching for variants with typos
Because none of this involves a focussed reading. It's the same reason why Level 3 can be less safe than Level 4. If you're skimming, you're less engaged than if you're reading in detail. (And if you're skipping around, you're missing context. You may catch Trump and Trup, but will you catch POTUD? Alternatively, if you just redact every mention of the President, you may wind up creating a President ***, thereby confirming what you were trying to redact.)
If it doesn't matter, automate it. If you care, have a team do a proper redaction.