User talk:CheMoBot/Q&A

If I were to try vandalising the CAS# on Boric acid, what would happen?

Q: Walkerma asked on IRC: "... If I were to try vandalising the CAS# on Boric acid, what would happen?"

A: What we have done with XLinkBot is that it reverts IP editors and new editors (accounts that were created less than 7 days ago). Individual editors can be whitelisted so that they are never reverted (that includes IPs and new users). What I want to do for now is to assign levels to the edits. For now:

0 = no edit to infobox (and no external link added)
>0 = edit to infobox or an external link added.

I have been working now with 6 levels, I envisage that levels 1-3 to be 'edits which can not be proved to be "bad"', 4-6 are edits which need to be checked.

Something like:

ALL changes to an editbox or which add an external link get at least a level 1, but higher when:

Changes:

Changing a watched field to another unverified value = 2
Setting a watched field which is forced unset = 4
Setting an unwatched field which is forced unset = 3
Setting a watched field which is set to '#NA' = 3
Setting an unwatched field which is set to '#NA' = 3
Setting a watched field to a value which is not correct = 5
Setting an unwatched field to a value which is 'not correct' = 3

Additions:

Adding a value to a watched field which is unset = 3
Adding a value to a watched field which is forced unset = 5
Adding a value to an unwatched field which is forced unset = 3
Adding a value to a watched field which is set to '#NA' = 3
Adding a wrong value to a watched field = 5
Adding a wrong value to an unwatched field = 3

Deletions:

Deleting a value from a watched field which was correct = 5
Deleting a value from an unwatched field which was correct = 3
Deleting a value from a watched field which was unset = 3
Deleting a value from a watched field which was forced unset = 3
Deleting a value from a watched field which was set to NA = 3
Deleting a value from a watched field which was not correct = 5
Deleting a value from a unwatched field which was not correct = 2

Also: Deleting a complete box = 6 Adding a complete (but 'wrong' box) = 6 (too high?) Change of box = 3

Here 'watched fields' are fields that we can 'control' (CASNo, mp, bp, etc.), 'unwatched fields' are the other fields in the box (other uses, etc.); 'forced unset' -> the verified box contains , meaning that there is no sensible value that can be put here, e.g. the refractive index of concrete; 'set to NA' means that the value is  (probably meaning that there is another field which contains this value, and this field schould not contain anything).

The range should now be in increasing 'severity', and everything above 3 should be at least noted, and probably repaired or reverted. I suggest in the end to revert new and ip users, for established editors it should be either flagged as 'wrong/unverified' in the page (with our colour coding), or repaired (that last may be controversial if a very established editor does find a wrong value in a verified box (which may happen)). In principle we can 'ignore' most of the things which are below 4, as those are in fields where (in wikipedia terms) there is no verified value (a simple change in the template, and these disappear from the logs).

I don't make a difference in severity between established, whitelisted or other users, it should affect how the edit is treated by the bot.

In the next phase (which requires a next BRFA) I would suggest it to mark the boxes where fields are not the same as the verified values, don't worry about reverting/repairing yet.

See pages under Wikipedia:WikiProject Chemicals/Log and Wikipedia:WikiProject Pharmacology/Log for some examples.

Thoughts? --Dirk Beetstra ^{T C} 16:32, 19 November 2008 (UTC)[reply]

Thank you for this - I'm really glad that you're documenting things. I have a feeling that if you disappeared tomorrow, we would not be able to manage the bot with out you, so if you can spare the time to do some documentation that would be great! Can you also define some of the basic terms, such as watched field, unwatched field, forced unset, etc?

As for my original question, I'm not sure I understand this well enough to know what it would mean! What I was wondering was, would the green text turn red? Or black? BTW, sorry I've been slow to respond, but I wanted to look at it properly before replying, and I've been very busy at work lately. Thanks, Walkerma (talk) 13:33, 22 November 2008 (UTC)[reply]

Hmm, that is what I try to get some grasp of. If we 'level' the edits into certain 'severity levels', then we can decide, per level, what we are going to do with it. In the beginning indeed, every change that makes a field verified or unverified should after say 5 minutes be followed by a bot-edit setting the appropriate fields to indicate the status of that field, resulting in the text turning green, red, or black. As I see it now, that should be for edits in level 4 and 5 (when 'wrong' data gets created; 6 being the removal of a complete box). --Dirk Beetstra ^{T C} 10:44, 24 November 2008 (UTC)[reply]

That sounds perfect! I'm guessing it would be red for levels 4 & 5, and then black for anything else? Or do levels below 4 not affect the actual data? Thanks, Walkerma (talk) 13:35, 24 November 2008 (UTC)[reply]

I think that 2 and 3 are things which are 'changing unverified data into other unverified data' .. but I see it also includes 'changing incorrect data to correct data' .. hmm .. this is more difficult then I thought. Maybe these levels don't help too much. --Dirk Beetstra ^{T C} 15:04, 24 November 2008 (UTC)[reply]

I'll be around on IRC today (Tuesday) at 1600h UTC for a bit, perhaps we can chat then? Walkerma (talk) 06:26, 25 November 2008 (UTC)[reply]