In an arena where plagiarism is harmful, I’d call this “negligence theory” rather than “mistake theory”. This isn’t just a misunderstanding or incorrect belief, it’s a sloppiness in research...
It’s sloppy, but my question is whether it’s unusually sloppy. That’s the difference between a mistake and negligence.
Compare this to car accidents. We expect that there’s an elevated proportion of “consistently unusually sloppy driving” among people at fault for causing car accidents relative to the general driving population. For example, if we look at the population of people who’ve been at fault for a car accident, we will find a higher-than-average level of drunk driving, texting while driving, tired driving, speeding, dangerous maneuvers, and so on.
However, we might also want to know the absolute proportion of at-fault drivers who are consistently unusually sloppy drivers, relative to those who are average or better-than-average drivers who had a “moment of sloppy driving” that happened to result in an accident.
As a toy example, imagine the population is:
1⁄4 consistently good drivers. As a population, they’re responsible for 5% of accidents.
1⁄2 average drivers. As a population, they’re responsible for 20% of accidents.
1⁄4 consistently bad drivers. As a population, they’re responsible for 75% of accidents.
In this toy example, good and average drivers are at fault for about 10% of all accidents.
When we see somebody commit an accident, this should, as you say, make us see them as substantially more likely to be a bad driver. It is also good incentives to punish this mistake in proportion to the damage done, the evidence about underlying factors (i.e. drunk driving), the remorse they display, and their previous driving record.
However, we should also bear in mind that there’s a low but nonzero chance that they’re not a bad driver, they just got unlucky.
Plagiarism interventions
Shifting back to plagiarism, the reason it can be useful to bear in mind that low-but-nonzero chance of a plagiarism “good faith error” is that it suggests interventions to lower the rate of that happening.
For example, I do all my research reading on my computer. Often, I copy/paste quotes from papers and PDFs into a file. What if there was a copy/paste feature that would also preserve the URL (from a browser) or the filename (from a downloaded PDF), and paste it in along with the text?
Alternatively, what if document editors had a feature where, when you copy/pasted text into them, they popped up an option to conveniently note the source or auto-format it as a quotation?
This is more speculative, but if we could digitize and make scientific papers and books publicly available, students could use plagiarism detectors as a precaution to make sure they haven’t committed “mistaken plagiarism” prior to publication. Failing to do so would be clearly negligent.
True—“harm reduction” is a tactic that helps with negligence or mistake, and less so with true adversarial situations. It’s worth remembering that improvements are improvements, even if only for some subset of infractions.
I don’t particularly worry about plagiarism very often—I’m not writing formal papers, but most of my internal documents benefit from a references appendix (or inline) for where data came from. I’d enjoy a plugin that does “referenceable copy/paste”, which includes the URL (or document title, or, for some things, a biblio-formatted source).
Negligence vs. mistake
It’s sloppy, but my question is whether it’s unusually sloppy. That’s the difference between a mistake and negligence.
Compare this to car accidents. We expect that there’s an elevated proportion of “consistently unusually sloppy driving” among people at fault for causing car accidents relative to the general driving population. For example, if we look at the population of people who’ve been at fault for a car accident, we will find a higher-than-average level of drunk driving, texting while driving, tired driving, speeding, dangerous maneuvers, and so on.
However, we might also want to know the absolute proportion of at-fault drivers who are consistently unusually sloppy drivers, relative to those who are average or better-than-average drivers who had a “moment of sloppy driving” that happened to result in an accident.
As a toy example, imagine the population is:
1⁄4 consistently good drivers. As a population, they’re responsible for 5% of accidents.
1⁄2 average drivers. As a population, they’re responsible for 20% of accidents.
1⁄4 consistently bad drivers. As a population, they’re responsible for 75% of accidents.
In this toy example, good and average drivers are at fault for about 10% of all accidents.
When we see somebody commit an accident, this should, as you say, make us see them as substantially more likely to be a bad driver. It is also good incentives to punish this mistake in proportion to the damage done, the evidence about underlying factors (i.e. drunk driving), the remorse they display, and their previous driving record.
However, we should also bear in mind that there’s a low but nonzero chance that they’re not a bad driver, they just got unlucky.
Plagiarism interventions
Shifting back to plagiarism, the reason it can be useful to bear in mind that low-but-nonzero chance of a plagiarism “good faith error” is that it suggests interventions to lower the rate of that happening.
For example, I do all my research reading on my computer. Often, I copy/paste quotes from papers and PDFs into a file. What if there was a copy/paste feature that would also preserve the URL (from a browser) or the filename (from a downloaded PDF), and paste it in along with the text?
Alternatively, what if document editors had a feature where, when you copy/pasted text into them, they popped up an option to conveniently note the source or auto-format it as a quotation?
This is more speculative, but if we could digitize and make scientific papers and books publicly available, students could use plagiarism detectors as a precaution to make sure they haven’t committed “mistaken plagiarism” prior to publication. Failing to do so would be clearly negligent.
True—“harm reduction” is a tactic that helps with negligence or mistake, and less so with true adversarial situations. It’s worth remembering that improvements are improvements, even if only for some subset of infractions.
I don’t particularly worry about plagiarism very often—I’m not writing formal papers, but most of my internal documents benefit from a references appendix (or inline) for where data came from. I’d enjoy a plugin that does “referenceable copy/paste”, which includes the URL (or document title, or, for some things, a biblio-formatted source).