Bot count, the sequel: 52%

Phil's assumption seems to be that none of these are true. All I'm saying is that they could be.
I don't think anything of the sort. I'm simply saying that neither Phil nor anyone else has provided any *evidence* to suggest that this isn't statistically significant, and suggested a couple of ways to test that hypothesis.

Phil asks why he should do that.

And I say that, because in the absence of better data, this is all we have to go on... and he obviously doesn't like the figures that came out of the study.

Correction Yosef; 52% of the population online are what Anya assumes are bots. We still haven't satisfactorily developed a way of telling who is a bot and who is just standing still because they are busy with other things. That's why Anya's data is meaningless at the moment.

If I were to do a survey, what would be the maximum bot percentage that I could report and not have people who don't know the real percentage tell me that I'm wrong?

ZERO or 1%

People here just want to argue for the sake of arguing and for whatever principle they are standing behind at the moment.

Anya's data is good in my book. But i suppose the rest of you busy invalidating it are recalling your statistics & analysis course in the process.

Agreed, Briana.

Anya's done this sort of survey before, and posted the results here.

Please notice that she is posting raw data, not drawing conclusions. Also, she has posted her data collection methods.

This is good, solid work. You can disagree with her methods, but please, if you do, also come up with some rationale as to why they are flawed or inaccurate or misleading, and suggest some better data collection scheme.

Is more data needed? Maybe we should assemble a Forum Bot Hunter team...say, ten volunteers who will conduct their own surveys. The sims to be surveyed could be chosen by some scheme, or selected at random. Each volunteer would cover, say, 100 sims.

If I thought for a second that LL would listen to the results, and do something useful with them, I'd volunteer to be part of such a study.

We still have no way of knowing if you're an argumentative trollbot coded in C++. So your opinion of Anya's data is meaningless at the moment.

Oh the irony.

Error 234 - u suck monkey balls.

brb, suicide.

More than 90 percent of people who commit suicide have a diagnosable mental disorder.

Rha, you are totally missing the point of Anya's studies, which is to investigate ways of *providing* numbers to back up the completely unsupported conjecture and speculation that is the order of the day in many arguments about bots.

There are points of Anya's methodology that can and should be discussed, but it is quite clear from Anya's points that she is well aware of the limitations of her methodology and qualifies her hypothesis and conclusions accordingly. She has even already responded to one such question about her methodology - namely, her initial choice to collect data during a high-concurrency time - by going back and collecting more data at a low-concurrency time.

Anya's work has the same value as any other attempt to glean information that is statistical in nature, be it electoral polls, FDA safety-and-efficacy studies, market research, or any of the other myriad areas in which people use small samples to test hypotheses about large populations. Anya's work also suffers from some of the same weaknesses as other statistical studies. But she hasn't claimed otherwise, and once again she's shown that she's willing to go out and collect more data to address concerns about her methodology.

That's what some people like to call the scientific method.

No one is taking her word as gospel, as you have accused. Some of us just don't see a reason to throw out her numbers or declare them worthless just because they don't answer every conceivable question one could have about the prevalence of bot-use gridwide.

And in particular, Rha, I truly cannot understand your apparent hostility toward Anya's work. You even suggest she might have fabricated the results which is, quite frankly, a bridge too far. Your escape hatch "I'm not saying she did" strikes me as a very lame attempt to avoid taking responsibility for such an outrageous accusation.

The problem we have here is that there is no way to truly collect any meaningful data.

For instance, I could walk up to a sim full of 20 people...and let's say it's a sex sim and everyone is IM, so no one is going to respond to my cheerful "hello, anyone want to talk?" Because no one responds, I (oops) assume that those 20 people are bots of one type or another.

Then let's say that I go to another sim and find 20 bots like the type I ran into the other day. Looks like an av, responds appropriately to some questions, but not to others. Because they respond, I have to assume (oops) that they are real people, and not bots.

Further, the bot programs typically have a chat function, where if you are sitting there, you can respond to questions. So, assume for the sake of argument, that Botowner A is sitting at the desk for a moment, when I walk by with my "hello, anyone want to talk?". He chats me up for a moment, and then wanders off to eat dinner and watch tv. Bot or av?

So, while Anya has done a cross-sampling, and I do applaud her for the time and effort, she produced data collected in a manner open to vastly different assumptions, and it therefore simply cannot be accepted as gospel, or even mildly conclusive.

That's all pretty much what I said.

OK, this is my final pass at it, because I don't think it is worth all that much effort. First, it isn't inaccurate since the data is self-defined and self-evaluated. A grant writer would make her correct the casual confluence of bots and campers (she indicates a total for bots and campers and defines the percentages as bots only). I admit that is pretty picky, especially considering my attitude about the issue, as opposed to the use of the data.

The primary flaw as raw data is that we do not know exactly how she defined a bot and we do not have someone else that used this methodology to do a similar survey. That is simply a fatal flaw. Period. It isn't even real data until the methodology is given and the conformity checked. If someone else tests it and finds that it produces clear, reproduceable events, then even flawed, it can probably be corrected and used in the future. Hasn't happened yet. No, I'm not going to do it myself.

In a discussion, it is further flawed, but I would have to argue on the either/or. She made no attempt to extrapolate, yet many people have. Correctly, she presented raw data and no assumptions (beyond how she determined bots). She simply said that on the sims surveyed, 52% were bots. If you take it no further, not one step further, the data is acceptable because it really says very little. It is also not misleading for the same reason (other than the camper/bot quibble).

If you want to use it for a meaningful survey, it is flawed for too many reasons. You really have to know the purpose, because if there is no purpose, then there is no reason for the exercise. If the purpose is simply to say that LL is wrong with their given percentages, then you need to establish how well the results extrapolate to the entire grid. Considering the structural differences between themed and covenanted islands and any mainland, that is a significant flaw. And what is misleading is when anyone uses this data to make that extrapolation.

I know that the purpose of the survey seems to be a non-item for most of you, but in grant writing you learn the significance. What IS the purpose of the count, other than to say "nanner, nanner, LL was wrong"? If the point is that the bots affect the SL experience (which I believe is true in many cases, but have no way to prove it), you would have to define your suppositions, your approach, and how your survey was going to prove THAT. IOW, IMO if you don't define why you are collecting data and how it would support change, what's the point? If the purpose it to show that bots affect performance, then you would have to determine what the impact of a bot had on performance incrementally and show how the count had relevance (if part of that metric was proximity of real avs, then you would have to be collecting not only numbers, but also at a minimum a proximity metric).

So, yeah, I think all it is is an exercise that produces numbers that have little relevance to anything other than disproving the LL numbers, and even there has no relevance to proving anything significant. Until you can start by telling me what the LL numbers would/should/could mean in the decision process about the future of SL in relevant, provable terms, "nanner, nanner" makes no difference. Just my opinion, based on rewriting a lot of grants in my time for this kind of stuff after the people that really know sent them back.

Anya also posted that she used profile data to help with her (admittedly objective) determination of what was a bot, and that when she was not certain she resolved the question conservatively by putting the av down as human. She noted that in crowded clubs it is difficult to tell bots from humans, but that she conservatively estimated all avs were human despite her own personal belief that some were not.

I am not saying that Anya's criteria are bullet-proof, only that there was more to it then just walking up to strangers and asking if they wanted to talk.

I would be very curious to hear some suggestions for other ways to determine whether an avatar is a bot or not. If we are very fortunate, perhaps Anya will be willing to conduct another study using refined criteria based upon such suggestions.

Here's a scenario for you. Dom John Doe and his sub Cleopatra are at a club. Cleo baby is definitely human, but has strict orders not to speak with anyone, so she does not. Assume for a moment that Cleo is an alt account, NPIOF (because John pays for everything), and was just created last week because it's the most convenient way for them to operate their RP. John chooses not to speak because he's enjoying his IM conversation with Cleo.

Again, you've got two people, one of whom "appears" to be a bot based on profile and payment info, but they're human. There are just far too many possibilities for flaws in the data.

I'm not going to keep arguing it, because to me, it's not worth fighting over. But nor am I going to accept the data at hand.

Jus sayin.

I think we can probably agree on one thing though; there's a whole lotta bots out there.

and some people in Second Life don't have any eyes and arms and so can't type. So it's quite possible Anya met some of those people. This would totally invalidate her numbers!

Jus' sayin

My tier is over 90,000 dolllars annually. Ostensibly for grid resources.

Bot runners pay... ah... nothing for bots!

Because why should they? I'm paying for the resources, and so is everyone owning or renting land... how cool of us!

* * * * *

I don't mind a bot or three per premium account, or free human usage. I can see the merit in that. But beyond that, get this abusive crap *off the grid* or pay for it, I say.

Hard to measure bots? Well, it will suddenly get easier if there is a financial protest. Or maybe we should sack traffic data - whatever it takes.

Didn't we just hear that resource over-use was the reason for charging openspaces into financial nosebleed-land?

Well gee, I know a *great* area where cash can be saved or made to help balance things... let's start with any avatar logged in 24/7, or anything that logs in that teleports 60+ times per hour, or a few other basics like that. And then add a few more. Don't like it? Mistaken identity? Have a chat with a human operator to get off the hook, and explain to them why you just did X land searches for the past 14 hours.

Then she should not call all campers bots.


308 / 591 = 52.1%.

Wanna say "52% probably as socially useless as a traffic bot"? Fine. Problem solved. (well, one problem solved)

We would have to ask Anya how she would mark down your hypothetical John and Cleo. Given what she's said in her posts, my guess is these would go in the "not sure and therefore mark as human" column. But that's my own favorable assumption about her methods, which she stated were conservative; maybe if she manages to slog this far into the thread she'll provide her own answer.

There are a lot of straw men being attacked on this thread. I think if you actually look at Anya's criteria and her claims some of the attacks evaporate. True, she may have marked some people as bots who were actually human, but she also likely marked many people as human who were actually bots. Sure, her studies don't provide definitive, unassailable, grid-wide numbers - after all, she only looked at a swath of mainland - but she's made some hypotheses based upon some reasoned guesses as how representative or not representative that swath might be. And she's shown the willingness to go back and ask further questions to try to address systematic errors in her analysis.

Once again, her work looks to me like a solid application of the scientific method. Definitive? No, an individual study of a statistical phenomenon rarely is. An informative and provocative start that sets the stage for more refined work by Anya and/or (let's hope "and" others in the future? Absolutely. And, in my opinion, heartily commendable.

I will not re-post the criteria every other post for the benefit of those not reading. You'd still find a way to not read it.



Of course they are. Most of the bots are in locations you will not happen across unless you are looking for them.



No, you won't.

I do not use non-responsiveness as a criteria. Did you think I talked to 700 avatars in ten hours?

I've tried it. I got nowhere.


"secks?"

" secks?"

"secks?"

"secks?"

"secks?"

"secks?"

"se...

Your scientific rigor is wanting.

The survey was done for the same sims and closely following the first so that only one variable changed, so I could investigate that variable: peak vs min concurrency.