Should Bots be used on Wikitree to modify profiles?

Question

It was suggested that we move this discussion to a new question, so... Should we use Bots to automate updates/modifications to profiles?  Does the benefit outweigh the problems caused (inadvertently).

Comments

Yes, (with all the previously discussed and thought out conditions testing, and approvals).  Think we need a Yes / No at the top of the post.  Comments/Answers below.

Thanks to Jonathan for all his efforts.

Answer 1

Bots might be OK for some limited uses as stated above, within certain parameters:

A collaborative effort for any bots created and used, maybe a Bot Project or something like that. 

Always open code, posted on the project or bot page.

If we are going to use them, let's build in some safeguards.  Like, more pairs of eyes in the first place, and step-wise approval process.

Limited Beta testing before any widespread use.

Maybe even an opt out option ... or at least restriction to Bot-appropriate profiles

Concensus on the appropriate uses

I almost hesitate to suggest this, because if done carelessly ... Yikes! but I personally would LOVE a GEDCCOM bomb clean-up bot, just to take all that Ancestry gibberish and put it in our source format, (instead of those awful 3-part <span> things) 

But that is a huge task, which would need NASA-level programming and fool-proof safe guards, many minds working on it, open code, repeat beta testing, small batch implementation, and very directed use (actually applies to any bot)

What I would not want to see is any one-writer, one-owner bots turned loose on WikiTree in general ... the very thought is horrifying.  This is the sort of thing that could go so very wrong.  Maybe this Forotree is our wake-up call ... if we learn this lesson now, then we probably got off cheaply.    it could have been much worse.

How many of you even knew this bot thing was happening? (I didn't)

Comments

The danger with opening up all the code is that it could then be used by someone with malicious intentions. The code for the ForoTree bot has been checked by others, although the code that caused the bug that has caused some upset was not checked and absolutely should have been. I take full responsibility for this and am very sorry as this was a huge oversight on my part.

---

It is my personal belief that this was released during the days just prior to and during the GFR to reduce the amount of comments like we are seeing in an effort to have them in place with a minimal amount of opposition.  I will always remain opposed to any bot that does anything other than possibly add a template or catagory listing to any profiles and even then in only very limited cases like the steps to AJ or GFR template where there is very little chance for error.  There is a possible problem with the opt out option and that is this, most of us share profile management with others.  If I did opt in and another manager of profiles that I am also on the watchlist for opposed their use the bot could still modify that profile against the wish of the other manager and that would violate the colloboration part of our honor code if my belief.  The only possible way to make this work is to have the bot only provide a suggestion with no way for automatic changes without having a verified human approve it,

---

There was no conspiracy. :-)

Jon mentioned in the WT_Apps e-mail list when he started working on this. It's a public list that anyone can join but it's really only for programmers.

It might have been a good idea for me to post on G2G that Jon was working on this. But honestly, if I posted every time some programmer said he was going to do something, we'd have a lot of posts on things that never happened.

I'm very impressed by Jon. He's creative, brilliant, shares our mission, and does what he says he's going to do. He's also been great about communicating and improving his code based on feedback.

---

Ditto Chris! I couldn't of said it better.

Answer 2

Perhaps we should be thinking in terms of what controls are needed so that a Bot can be used to make needed changes to profiles.

All of us make typing errors.  We have set up many categorees that for various reasons need to be changed.  Some of these changes require hundreds or thousands of profiles to be accessed and fixed.

If we can get an automated process of doing the work - that would be wonderful.  But we need to be sure that there is a process in place so the the fixes do not create even more errors.

Comments

Phillip, There is another question that asks if there should be such an emphasis on catagories in the first place.

---

Dale,

The emphais on categories is I, hope, a temporary thing.  The Categorization Proejct has been working overtime to clean up problems that most of us did not even know existed.

Did you really understand that a green category may not be set up at all?  I did not and I am fairly certain that almost all other managers on Wikitree did not either.

Another month or six weeks we will probably have things well in control and be able to be in a more reasonable maintenance mode.

Answer 3

As I understand it -- it is limited to your "watchlist" and has to be added as "trusted" in order for it to work.

It appears there are only 3 uses at this time:

FamilySearch Citation Cleaner Upper Bot

check all your profiles for FamilySearch citations and move the bloat into your sources section rather than in the main part of your profile, making your profiles much easier to edit.

Global Family Reunion Bot

check how many steps away you are from A.J. Jacobs (the organiser of the Global Family Reunion), and then add the "Global Family Reunion" template to the top of that profile.

UTF-8 Encoding Correction Bot

correct what it thought were errors with special characters.  Probably not needed except on case by case basis.

You can read about it on Jonathan's forum and the suggestions currently being made.

I believe that any "bots" used would have to be formally approved first before being released.

My opinion only is that it could be a useful tool, especially for mind numbing tasks and especially in cleanup of those profiles that have ancestry/Gedcom import issues.  It seems like the majority of these do not have any "real source" outside of a "tree" which only members of Ancestry can access -- so you don't really know what is there.  I suspect that it could be used to "conform" this information with one blanket statement and maintain the link or links at the bottom in a category of its own, deleting the redundant info that is now being done manually.  Or, perhaps tagging with a category of "Unsourced - Gedcom".

Just my thoughts.

Comments

Good post, Sandy.

---

Ditto Chris!

I agree the bots should be approved. On the other hand a small group of people need to test these bots to see how they work. I've been helping Jon  test these bots and watching the changes closely made to my profiles, At least, the changes are contained within my watchlist making it easily to spot bugs.

Having a small group of testers helps narrow down any issues and making corrections before releasing the bot to everyone.

I like your tagging idea Sandy. ( using my phone so I'd like to post more about this later)

Answer 4

I do think there can be good uses of automation on WikiTree. In fact, automation already being used behind the scene by the WikiTree team to pre-calculate relationships with AJ and find DNA relatives for instance.

I have also used some automation myself in the form of a Chrome extension to help locate illogical dates in profiles. My extension doesn't try to make changes, it is only meant to alert users of problem areas.

I have thought of potential tools that could automatically edit profiles to fix stuff, but I haven't dared to try something like that yet because with great power comes great responsibility. I think that someone that does write a bot that can edit profiles should do it very carefully, and have a robust plan for testing and perhapse also have the code reviewed by others before deploying it for all to use.

Perhapse tools that are meant to edit data could be setup in a way where they need to operate with a "human in the loop". In other words, it may present the user with the edit page with the changes made, and leave it to the user to double check that the edits are correct, and actually click the submit button.

Comments

Editing profiles by automated means is opening the door to more problems like have already happened over the weekend, and the AJ connection has to initiated by a human it is not automatic,

---

Good post, Roland.

We've been running "Match Bot" for years.

The value of bots on a wiki, I think, is pretty well-established. We should probably be looking to Wikipedia's usage of bots for guidance.

Answer 5

I think for many of us, someone needs to explain just what kinds of changes you are talking about and how would those changes be controlled?

Comments

Robin,

I agree with all Roland's comments regarding very careful review and test BEFORE implementation on any kind of scale, no matter what kind of tasks are undertaken by an automated tool of any sort.

I strongly believe that some editing tasks should never be automated because the decisions on whether or not to make a change in different specific instances will not always be the same.  I believe that all of us have often felt bogged down by all the seemingly repetitive typing needed to enter certain kinds of information on profiles, but I also believe that the greater portion of these tasks are not appropriate candidates for automation.

I am also a very strong proponent of utilizing technological advancements to our benefit whenever feasible, however.  I spent my career on what has been called "the bleeding edge of technology", where I was continuously challenged to find ways to gain advantages from every new development.  It is a very exciting place to be and it is very difficult for me to say a firm "no" to application of new capabilities that enhance editing efficiency as impressively as bots can, however in a situation where there is inherent risk of data corruption there is compelling reason for doing so.

I believe that we can make intelligent decisions to selectively designate specific tasks for which the capabilities offered by Jonathan's bots would be a wonderful tool for us to have.  These would be situations in which risk of data corruption is low and ability to mitigate the risk is high.  

Perhaps this was not rolled out in the best way possible and perhaps we need to put the brakes on it - hard - until we have the opportunity to consider all the potential uses for bots and select just one that does not carry risk of irreparable data corruption.  We could then put it in use for an evaluation period, after which we could asses its value.  If all goes well, we might then expand the use of bots to a very carefiully selected few additional tasks, followed by an assessment.  After that, every time someone has an idea for an additional task that would be suitable for a bot,, consideration of the potential benefits, weighed against a full risk analysis should be performed before any decision is made to idevelop and implement the new bot.

I would like to propose a task that I think meets the critera for the initial evaluation.  The categorization project is facing several monumental tasks right now.  Several category names need to be changed.  In many cases, there are large numbers of profiles in a category whose name is being changed and it will entail a very large effort to manually go into each of these profiles and change the category tag.  I propose that the categorization project select ONLY one category whose name is being changed and ask Jonathan to create a bot that will make that change.  After that task has been completed, if it is successful then we can use it for the rest of the category name changes.

Robin, I absolutely agree on the need for controls.  I would like to see a small group - between 3 and 6 members who are computer savvy - designated as the evaluation committee.  They would look at the changes made by the bot and make the decision as to whether bot use should be expanded every time new tasks are proposed for it.

---

I completely agree with Gaile.  Bots have the potential to be a great help, and to do great harm.  There should be a stringent evaluation before they are released for use.

Answer 6

I see no benefit to using them, and yes I did check them out.  Having any automated system change profile biography or sources means that we will no longer have any control over what is true or even personal preference over the style of source information that we have on here.  With the debate about sources currently raging about source formatting how could a bot have already be written for the purpose of "cleaning up" "bloat"? This discussion should have taken place prior to the start of this program rather than after it has already caused problems with not only sources but with even names when using non English characters.

Comments

I could not agree more about the unsuitability of using a bot to automatically change source formatting - no matter what the outcome of the current discussion on style guide sanction of one or more specific source format conventions.

This does not mean, though, that bots do not have any place here for anything - there are tasks for which they are potentially very suitable and, in those cases, would save uncountable hours of tedious human labor.  As I said in a few other places, it requires very careful analysis to reach an intelligent decision as to when to empoy a bot for any specific task.

---

The use of bots does not necessarily mean we lose control. We can always go back in and manually correct  or change something.

As for "personal preferences," a wiki is not the place for them. If I want my personal preferences to rule, I use a privately edited tree elsewhere.

---

Jilliane, I,for one, will be spending most of my time and effort on my privately edited tree as you mentioned.  I have not thought much about totally leaving WikiTree yet but I will be doing a lot less on here and will probably wind up like a lot of others just fading away.

---

Dale,

I've had to adjust my expectations a few times here. The wiki-nature of this system, to me, anyway, means that if I want to play here, I need to play by the rules created by the community and/or the people who run the system.

When I disagree with the rules, I will make a case for changing them.  Sometimes I make a case successfully; sometimes not. Each time "not" happens, I reassess to what degree I invest here.  

I'm still here. But I have changed both the amount of time spent and the types of things I work on. It's life.

---

I am still here, for now, but with all of the push for more projects, catagories, connection to AJ, and now bots I am going to do more source work on my personal family tree and less here.  I have already asked to be removed from two projects that I was on and will be requwsting to be removed from the remaining two soon.  I already downloaded a GEDCOM of my wachlist and I am reducing it's size.  Any new work will be primarily in an offline site or possibly with a competing site.

---

By reshowing my answers and comments Chris is fanning the flames.  I wish things were different but by doing this he is showing no compassion to my effort to
Dont Wiki When Angry so this will be one of the very last itmes you see my name on G2G.

Answer 7

Generally speaking bots do not have a good reputation, they are the ones responsible for the captcha requirement on many website (some bots find register form and randomly create false users) or the ones who score the web finding unprotected email address for spamming.

But a bot (like this sentence "but a bot") properly design can save a lot of tedious task, it work similarly as the find and replace function in many good text editor, if you have a page or two to do it's not a big deal but if you count them in the hundreds, quite a different story.

Comments

And for those of us who are trying to help improve a couple of thousand profiles the ocurance of a "minor" error on the profiles could result in a lot of hours fixing those errors, IF WE CATCH THEM.

---

Dale,

I fully agree on the IF WE CATCH THEM part, but, when caught, the error can be repaired by a bot, so that it would not result in any burden on people to do the correction.

Of course, this is predicated on the requirement that specific instances selected for bot usage are carefully selected with full consideration of the potential for errors, a full blown risk analysis of the impact if any errors do occur, and concurrence by a group of people who are technically knowledgeable.  With that kind of system in place to control the use of bots, I believe they have the potential to be a very valuable tool for us.

---

The reality is that in this last problem he did try to use a bot to fix it but there were still some that had to fixed by a human. I would expect a programmer to defend bots but the facts are clear this bot caused a problem that even another bot could not completely fix.

---

It is not true to say a bot couldn't fix it, only that it wasn't programmed to fix all cases on this occasion. On this occassion, due to the relatively small number of profiles affected, it was more efficient for me to fix it manually rather than write a not to do so. Had it affected more profiles, it would have been possible to write a not to do it. The bot actually did a very efficient job of highlighting which profiles needed manually fixing. Also, it should probably be noted that this bug in the not was partly due to a bug in WikiTree which I misunderstood. I've posted about this in a new discussion thread.

---

On one of your answers or comments about the problems YOU sait that in the few cases the bot could not fix the problem you restored the profile, your words not ours.  See your own answer on http://www.wikitree.com/g2g/154631/what-is-botforotree

---

There's a long history of bot usage on Wikipedia. We should probably be looking to http://en.wikipedia.org/wiki/Wikipedia:Bot_policy to see the lessons learned.

Answer 8

NO........

Comments

Why would Vincent's response be flagged?

---

The other question about bots from the othrer day was flaged and hidden.  It has since been reopened but I suspect that someone is trying to remove the opposition to bots from these threads.

---

I remember when I first joined Wikitree I think I flagged a question thinking it represented a Star or a positive agreement.  Maybe it's something like that?,,,,,,

---

I didn't flag it but a simple NO on a question like this does deserve some kind of a smack.

NO... because?

---

Claude, I disagree.  What part of NO don't you understand?  It answered the question "Should Bots be used?"

---

Sorry I didn't know it was a poll, oh! wait a minute, it is a discussion board, what part of discussion don't you understand?

---

I agree  with Vincent's answer (but a bit louder) - NO

And just so people know, when I saw the flag the other day, it was me that hit the second flag to hide the question. It's a blatant commercial and unauthorised intrusion into the Wikitree system.

---

Peter I agree with most of your comment but since Chris thanked Jonathan for creating the bot I can not saay that it was unauthorized and by hiding the discussion you only allowed things to be more confused.

---

Thanks Dale,

I was not aware that Chris had given tacit approval to Jonathan to introduce this.

Nevertheless my concerns remain. It was said a day or two ago that there should at least be a cost/benefit analysis before introduction of this sort of technology, I fear the costs will (at present) far outweigh any benefits.

---

Thanks, guys.

Jon has been discussing this in the WT_Apps group. It's a public e-mail list but it's only intended for programmers with an interest in helping WikiTree.

He mentioned many weeks ago that he was going to work on bots. I encouraged him since bots have been used to good effect on Wikipedia and we'd always wanted to develop more of them (in addition to Match Bot).

Honestly, I didn't expect him to produce anything. Over the years many programmers have offered to create apps that would help WikiTree members and our mission, but very few ever follow through. (Roland is one of the few.)

Last week, Jon announced to the group what he had created. I was blown away by it. Obviously, there are great risks and problems, but there's also great promise.

Jon should be encouraged. His talent and initiative are rare and valuable.

Answer 9

Sounds like venturing into Ancestry territory here. Ancestry has tried bot type sequences over the years - usually with disastrous results that many of us are still trying to get corrected in various trees.

It may mean more work - but in the long run - a human making a correction, a change or pointing out a problem cannot be beaten.

Comments

Thanks, Jackie. WikiTree is fundamentally based on human contributions. That should never change.

Bots should follow the model of Wikipedia bots http://en.wikipedia.org/wiki/Wikipedia:Bot_policy

Answer 10

I feel the need to weigh in on this.

Granted I am not as well educated as many and I do not have the degrees and college learning.

What I know of Genealogy is mainly self-taught I love to read and spend much of my free time doing so. I have asked others how to do things such as sourcing.

I have been reading about the bot subject and the input from others. Even though I understand their point of view I do not like the idea of a bot working on profiles.  Before you down vote let me explain.

One of the reasons Wiki tree works so well is the human touch. We work with each other and help each other. Wikitree basis is to create profiles for OUR family. In order to do this we need to stay with the human touch.

If we start with the bots where is the line drawn? Many of us once we add all we can to a profile orphan the profile for others to work on or hopefully a direct descendant to see and adopt. Will these now be adopted and a bot installed to manage it? Will a bot have the little piece of each of us we all leave behind on a profile we work on? Will a bot understand the bread crumbs some of us leave on a profile when we cannot go any further with it?

Will it come to the point that all someone has to do is upload a gedcom and then set a bot and forget about it? We have all seen where just uploading gedcoms without the touches now employed by the forest elf and others have led us.

Like I said one the reason we are such a great site is we as a FAMILY work together.

 

Comments

Great post, Jacky.

Interestingly, my comment to Jackie above could have been posted here in reply to Jacky. :-)

There's no doubt we need to be very careful in this area, but I don't think we should reject all automated edits. Match Bot has been running for years and although it makes dumb mistakes I think most people would agree it's been beneficial.

Answer 11

A bot might be useful after careful evaluation in changing categories. I do not like the idea of changing or reformatting sources or editing text. I also do not think that limiting the use of the bot to those profiles on one person's watchlist is an adequate protection. Many of the profiles on my watchlist are shared with others, either trusted list or co-managers. Are you going to get the consent of everyone on a trusted list for one profile before the bot is used? How would that work?

Comments

Currently the bot doesn't check whether there are multiple profile managers, but this check could be added relatively easily and those profiles could be skipped. I do think there is a case here for reminding profile managers that they do not get to discharge all their responsibilities when they use a bot, even if the not is following all the appropriate guidelines. That being said, it seems it is usually the bot that is blamed and so it is the responsibility of the developer to encourage this as much as possible.

Answer 12

A bot that would notice empty birth fields and place some sort of maintenance category on them would be nice.

Answer 13

IMHO I feel like bots are against the whole idea of wikitree as it was designed to be a collaborative effort by the global population of people to create a single global family tree. A 'bot' is not a person and as such should not be making changes.  There is too much potential for upsetting real people by using this methodology and it just seems pretty straightforward that real people should be editing profiles for other real people (alive or not).  Just my two cents.  And I've not been impressed with the changes I have seen made by the bots.

The only exception I could see to this would be having a bot that removes all blank profiles--those with a name and no dates or places whatsoever and that are not linked to anyone else at all.  I've been seeing some of these of late and as they are, in reality, unable to be even begun to be sourced due to no info to start with, having them removed would be productive.

But again, having a non human make that decision, might yet, be inappropriate to the 'people collaborating' concept behind wikitree.  Yep, probably so, best left alone.  LOL  I'll keep working em as I see em and as more folks join wt we will have more help!  So there's my final answer...RECRUITE MORE SERIOUS WIKITREERS!!  :D

Answer 14

Bots should be one-time ad-hoc runs instigated by a specific user at a specific time, and any changes they make should be done under that user's ID. Anything else throws accountability out the window.