Potential solutions to stop, restrict or disincentivize bots & alternatives for better bots - ๐ค BOTS! ๐ Act 3
In the previous post, Act 2 in the series, I asked "Are bots bad for Steemit?". Several arguments for and against were discussed and the author concluded that bots were not necessarily a force for evil.
This post will show that the Steem blockchain must allow bots, to maintain privacy and anti-censorship features. But it's not as bad as it sounds, and I will demonstrate why.
I make some suggestions for possible ideas for restricting bots, showing why some don't work, and others do in a limited way.
TLDR
No censorship ๐ซ ๐ถ & equality of access ๐ ๐บ = open API ๐ = bots ๐ค. Close API to kill bots but also kill platform. Tweaks could disincentivize bots while keeping API open. Or we could make the social cost of running a bot too high
Fixing the bot problem - practical solutions
Note, this article assumes there actually is a bot problem. That is a problem is not a given but is roundly debated in previous posts in this series.
I've spent some time thinking out several scenarios which could stop, restrict or disincentivize bots accessing the Steem blockchain. Some of them are ideas I've seen in posts around, some are my own ideas.
Traditional solution 1 & 2 are approaches you find on many websites and online services.
Hard fork solution is something that would require a change to the blockchain without restricting API access. This is the most effective area of solutions but the most controversial.
Voluntary solution is an example of an idea that works by voluntary agreement of the community within the current system we have. They could only be encouraged, but could be enforced informally as communities often do by shaming, boycotting and ostracization.
Foreward on solutions
API stands for Application Programming Interface, and is a very widely used and abused term in computer systems. In this usage, the operative term is Interface, as it supplies the means of connecting, i.e. interfacing, with the blockchain. Many web services have APIs available to access their systems, such as Twitter, Google Maps, eBay, Bitcoin charts, and many many many more.
The Steem API is the means for any connection to the Steem blockchain (including bots) for anyone who is not a witness. Witnesses run nodes which (in general) supply a public facing API connection. In this way, Steem is decentralized as there is more than one point of connection.
The API allows anyone to access the blockchain, given they have the credentials to do so. In other words, anyone with an account. This is important: all accounts have the exact same level of credentials, i.e. there are no privileged users, no admins or moderators. This means that the account you have, that my bot has, that @abit has and even that @ned has, they are all equally credentialed. The only things that differentiates us are Steem Power and reputation which (theoretically) anyone can alter by buying / earning more Steem Power, and community approval boosting ones rep, respectively.
It is important to know this. It shows that essentially bots have the same rights as people on Steem. If we reach the singularity sooner than I think, this means we will be the first (ro)bot friendly social network in the world! ๐
Traditional solution 1: no API
Do not allow an API, simple as that. All access to the system must be through an official portal, such as Steemit.com
The glaring downside is that this is patently against the aims of Steemit. Since the blockchain exists between witness nodes (similarly to miner nodes on most other blockchains, Bitcoin included) those who do not run one can only access the blockchains via a witness node. Of course it's not practical for every Steem user to run a witness, not to mention the crushing onboarding problem this would pose if mandatory. That is why the API exists, to allow connection to any node, thus leveraging the advantages of a distributed system.
Any change to this would remove the advantages and run the risk of censorship of the blockchain. This censorship problem is a fundamental one which Steem and Steemit are designed to solve, it would be a 180 degree turn to implement it. From the white paper Censorship section:
Individual websites such as steemit.com may censor content on their particular site, but content published on the blockchain is inherently broadcast traffic and mirrors all around the world may continue to make it available.
Steem witness code could be changed to mandate that Steemit.com is the only portal to the blockchain and implement the usual bot busting techniques (captcha, etc) and disable API (or limit it, but more on that below).
It would be very hard to get the witness to agree to it because this proposal is absurd in the context of Steem and against the core philosophy. I present it here not as a straw man but because it is such a common approach (perhaps the most common approach) on the web, and some people suggest this approach from time to time.
In any case the system is already in the wild, all the code is there, so it's too late for that. ๐
Traditional solution 2: restrict API access
You will be hard pressed to find a modern public API available today that allows free and open access (you can search the largest directory in the world here for them). Almost all require the use of an API key to authenticate access. Even when access is free, this usually amounts to getting a free API key and a required registration of some details, including at least an email address.
Steem uses tons of keys, surely one of them must be the API? Sort of but not really. When you sign up to create an app on Twitter (i.e. which uses their API to send and read tweets, etc) you register as a developer and are given an API key. This API key is required in addition to user authorization.
I should note that user passwords etc are not generally handled by third party developers any more in modern 3rd party systems, authorization is given by OAuth style logins, similar to Steemit Connect.
Currently on Steem, your private posting key acts as both posting and voting authorization for the associated account, and Steem API authorization. This effectively gives every user account the possibility to be a third party developer, which is exactly what a bot is - third party software running "on" the blockchain via a user account.
Like the Traditional solution 1, this is required by the blockchain because nobody can act on behalf of another user as the user with the notable exception of proxy witness voting and the HF17 proposed proxy post voting. It is not possible for a trusted party to post on your behalf, your key is the only authorization. It is only possible for someone else to use you key and post as you.
The distinction is important because it means that changing it would require weakening the security of our keys. For the same censorship resistance reason as the first solution, it would not fly.
Hard fork solution
One idea I had is that voting power rate limiting could be linked to time since last post. Rate limiting already has an effect on votes, what about extending it?
Rate limited voting, is the mechanism which causes you to loose voting power each time you vote. This is what encourages more carefully considered voting and actually is an explicitly anti-bot measure. From Steem white paper Rate Limited Voting section:
Individual users can only read and evaluate so many work items per day. Any attempt to vote more frequently than this is a sign of automation and potential abuse.
Through rate limiting, stakeholders who vote more frequently have each vote count for less than stakeholders who vote less frequently.
What about if those that post less frequently were to have each vote count for less?
To this I would add one controversial feature: any account with minus reputation should have their votes account for exactly zero. Losing reputation should have more meaning. This coupled with the above measure would allow the community to decide that an account was abusive and should not be permitted to vote.
In my view this would solve two problems:
- The engagement problem - those who do not write posts should not be permitted to vote meaningfully, this is an incentive to write and engage with other authors by commenting
- Any voting bot account would be forced to post in order for them to operate. @asshole would not have been able to exploit the system if this were implemented.
An average, active human user probably posts a few times a week (but it would be good to get actual numbers on this). Posting once a month would be the limit to qualifying as "active". Voting power recovery rate could be adjusted to slow increasingly in proportion to the length of time since the user's last post.
This would make "buying in" no longer a way to have influence on Steem by itself alone to have a voting effect, including the ability to reap curation rewards. It would no longer justified simply by capital but also by community engagement (last post time) and community approval (rep).
I can imagine there being wide support among users for a change like this, though I'm not so sure about witnesses. It mandates social network activity for voting power to have any use, both for personal reward and benevolent pool distribution influence. It would effectively disable the votes of silent whales who contribute nothing to the social network right now, though their previous actions (up to and including buying Steem with US dollars or BTC) would still have positive effect on the health of the Steem project as a whole.
Implementing this would also mean that any account which has not yet made a post will of course have infinite time since their last post and so must make at least one post to get meaningfully involved in voting.
You might say, well they can just automate a post once a day to say "Hey, still here." That's where the interaction with reputation comes in. Because reputation can only be effected by down voting posts, the more a user is encouraged to post, the more one has to judge them on. However this point was the main contention when I did put this to a few witnesses on # witness channel in steemit.chat. Some felt that it would create spam. I'm still not convinced of this, one post a month is hardly a spam problem.
Perhaps a more troublesome implication is that what this solution implies is that Steemit should be an authors platform. There was no real agreement that this is the case. Another contention was pro-freedom, i.e. forcing posting is restricting of behavior. But it should be noted that one is only forced if one wants their stake to be of use for voting and curation rewards.
It also denies the roll of curators who are only curators. I think it was an original idea of Steem to allow for accounts who do not post but are only curators, but I think what we've seen is that this account form is not beneficial to the platform and gives bots too much of an advantage.
I would really like to debate this idea so please comment if you have any thoughts.
Voluntary solution: Self-limiting
There's a strong argument that whatever can be done, will be done. If bots are allowed they will certainly be used.
However as a community we can make the use of them so stigmatized that the benefit of using them drops below the threshold of action. This kind of repression is what we humans excel at, and in fact that goes for other primates too.
Since there is currently significant support or at least toleration for bot usage, there would need to be some wide scale convincing. This political challenge may be achieved but I have not idea how.
It could be as simple as appealing to the assumed innate decency of people, if one was able to frame it in such terms convincingly and reach a receptive audience.
A less nice solution would be a kind of "scorched earth" approach, as spectacularly exemplified by the @asshole account, and less so by other accounts which I will not name here (though I have my own blacklist). This black hat hacker style, while not widely appreciated, is effective where nuanced argument fails.
The advantage of competitions of bots, or an appeal to ethics
On a system which permits bots, restricting usage of them is one way to go. But for some who either disagree or are tempted by the rewards, this will never affect them and they will continue to use bots.
A solution is to provide alternative bots which appeal to ethical usage of users. Here are two such ideas:
Read-only bots
On my last bots post, @edje said in detail what many people feel, that bots should not vote (or presumably comment either) but may be useful for other things. I call these bots "read-only", i.e. they do not replace user action.
Here's some of what he said:
I think bots can be a tool to help filtering the number of posts based on a set of parameters for the human individual to have a filtered list to go through manually. It may help to identify copied posts, it may help to identify undervalued posts etc. Essentially I see the bot as an good bot when it helps analysing the posts and comments made on Steemit. But thats were I draw the line.
I agree, what we need are good filters, there is no "organic" discovery in a sea of information, it all must be processed in some way. Contrary to the mainstream approach of proprietary and secret algorithms, open and free choice in filtering is what's needed.
I had already made a project idea for this, to separate analysis from voting with my own bot, and I think in the next months it will be something I'll work on as there seems to be interest in that kind of thing.
Custom post filtering and notifications could even be added to steemit.com, either officially or unofficially if they open up a plugin / 3rd party app system (as some have called for here and here).
I predict that this is the internet we will see more of, the IoF or "internet of filters". We're already there, Google's search algorithm and Facebook s news feed do it, but we need it to be customisable and the filters transparent, mutable (changeable), free and applicable to our data and the data we interact with
This will help us to find the stuff great votes are made of, and make the decision ourselves. Some people will still choose to vote automatically if it makes sense to do so, but supporting an alternative view on voting ethics sounds good to me too.
Vote prompting bots
I've been impressed with the eSteem mobile app for Android by @good-karma and in particular I'm really happy to get mention notifications on my phone.
It give me the idea that, following on from the idea of read-only bots, one could get a notification of a new post which has been chosen by your custom filter bots, and then prompt you to read the post and optionally vote on it.
This to me really seems like the best of both worlds: quick voting on great posts but after reading. Of course no one can force you to read but this would encourage more people who want to use bots to make the choice to be more engaged.
Finally
Thanks for reading! This has been the third instalment of my bots series, looking at whether bots are bad for Steem or not.
Next in the series will be an examination of the kinds of bots which I am aware of which exist on Steem. If you have one of your own, please mention it in the comments with a link to help my research! ๐
I agree with your proposal about requiring posting to enable voting. There has been more than a few instances of non-posting high SP accounts flagging, and I have felt from very early on this was not beneficial.
My version of this proposal goes a step further and has a decay rate for reputation, which then enforces continued engagement to have a high rep, and added to the current system, plus your idea of requiring activity to be able to vote, would result in a more careful use of SP and a more clear and effective application of the reputation score to reward contributors and punish those who take away from the platform only.
I have also suggested that witnesses be more constricted in regards to their engagement, that is, as well as votes, the reputation of the witness account factors against their position in the witness leaderboard. There is some objections to that in that sometimes witnesses are too busy doing witnessy things to make posts, but I think if you take your idea, of making 'active account' a prerequisite, you can use this decay curve to enforce it. The decay curve would then also have a decay rate that accelerates over time, after 1 week, maybe you lose 1%, the next, maybe 4%, maybe 3 weeks, 10%, at 4 weeks you lose 25% of the effect of your votes and at 5 weeks your effect is down to 50%.
This would completely stop the use of big stake to censor users because they would have to subject themselves also to the judgement of others by posting.
Although a long post, you created a very good post explaining what is possible and what is not possible. That I think should give more insights to those readers who do not know how Steem and Steemit are build. At least it gave me a good insight. BTW, thank you for mentioning me and including a quote from my comment to your previous post on bots. It gives me a good feeling, my contribution is appreciated and actually used :)
Regarding this post, I'm not sure if I can add something other than that I think indeed that as much as possible we need to technical enforce rules rather than by culture. Before implementing changes, we may try to align the community in acting as if we would be enforced by technology, in order to get practical field experience and determine if the results are indeed what we would expect and determine if the results are good (or not) for Steemit as a service and more importantly, for the community and its envisioned growth.
I repeat myself, when I state that I opt for read only bots, with some exceptions if that is possible; The exceptions are for instance @cheetah and the twitter bot placing a comment. But bots shall not be able to vote indeed. How this can be implemented is beyond my expertise. Since my knowledge on how exactly reputation is effected, I cannot comment on the proposed implementation. The comment of @l0k1 looks valid to me, but also here I think I have too little knowledge to understand it completely. I hope other readers and commenters are more equipped to understand what a good approach is and what the consequences are.
I thought what you were saying was in line with what a large number of people think and it was great chatting to you, thanks for the quote ๐
This is certainly what the creators thought. You're probably right, there is probably never going to be a completely dominant culture here, and if there is there will always be a minority who can significantly disrupt things.
Leading from this, there can be no exceptions with this technology, no privileged bots such as @cheetah. All must play by the same rules. It makes it really challenging to make rules which fit everyone and everything, and why the best witness are always calling for KISS - Keep It Simple Stupid!
I really don't think bots can be prevented from voting, and there are good reasons to allow it. I think people who object to that usage (like you do) should focus on disencentivizing that particular behavior and / or the advantage bots have over humans. That's really the point of the post. The vote rate limiting feature is already something that does that. My Hard fork solution above would it. More ideas for this are needed and to get the devs and witnesses on board. Not easy but I think there would be a lot of general user support, as I said in the post.
Thanks again for your comment and engaging with me in conversation on this topic ๐ ๐
Lets indeed assume that we stay/create an environment where bots can play a role, I would try and still stick to the rule bots:
When allowing bots to place posts, this can be used for external created content to inject into Steemit as well, think of ezine using Steemit as just another channel of interacting with their followers.
When allowing bots top place comments, this helps bots like @cheetah, or anybody who wants to leave comments when upvoting or even downvoting content, manually, reducing the work required in curation and securing Steemit (especially for downvoting activities).
I would not allow bots to vote though. Let voting always be a manual activity.
I have no idea if this can be technically created, but that would be the model I would opt for.
Can you please re-read the post? I talk about how we cannot have what you are proposing because there is no distinction made between human and bot accounts.
That somehow I forgot again, thanks for reminding!
That makes it indeed a complex topic. Since not voting for bots, means no voting for humans. I suggested some weeks ago to look into behavioural analyses. Through such analyses find the bots and block them. In the telecommunication space, and in particular in the SMS space, they try to solve a problem they have with advertisers using backdoors to get there push SMS into the telcos networks. These backdoors are either free of charge, or offered to them for a very very low price. Systems are created looking at logging data of SMS senders and try to figure out who are these backdoors users and block them when identified. This may not be the next step for Steemit to implement, but could be step 2. First step is indeed to de-incentify the bots as you are trying to get information and discussions going through this post.
I think there are some very simple markers for bot activity that could be used because bots behaviour is usually very repetitive, periodic and / or does large batch jobs that would be impossible for a human.
I have marked an idea to create a simple web app to detect this stuff which I'd like to do some weekend I have free. Of course the great irony is that this would actually be a bot!
There is no official stance against bots for Steemit Inc., I think they in fact want them, so I would not expect any systematic de-bot-ifying from them. It's only some (maybe a lot of) users which are against them.
As you know, I'm not against bots, and as you state, the community as a whole, or the witnesses and Steemit Inc as a subset, may not want to force bots out of the network. Analysing bots, ie read only as you name them, are very good I think.
The effect of blocking bots, or reducing their influence when used in a 'bad' way, will result into more intelligent bots, ie trying to circumvent the bot detection algorithms. The good thing from this is that it'll create more intelligent bots. The 'bad' thing is that it requires constant work on the analysing bots. But, this gives good things again, since more intelligent bots may be applied to other social networks, or big data environment which gives a business opportunity and revenue opportunity for those who actually create the technology.
If @l0k1 's proposal is implemented or something similar, it would be possible for those disagreeing with bots or certain kinds of bots to have more power to oppose them. But then again, under his proposal it would cost something for the opposition. ๐
If such a situation were to arise and bots were to have to adapt to avoid detection, it would mean they would have to act more "human-like", which would by definition reduce their advantage and impact to the level of human, assuming there are no exploitable loopholes (there almost certainly will be, but they might be small). That would be a significant win for the bot opposition!
And you're right, and pressure to innovate could have net positive effect.
Just before I read your last comment to this threat, I read the entire threat you had with @l0k1. Very interesting proposal made indeed.
you do realize that steemit.com is basically a webapp that does pretty much everything it does via the open API? Just right-click in chrome and click "Inspect" for the Developer Tools window and you'll see all the SteemApi calls. If you try to block bots, you block steemit too.
Yes I do! This post is pretty much that exact statement.
Maybe I'm missing something in your post, but that would make the above statement impossible to implement successfully. Any bot could easily talk through the API the same way steemit.com does. And even if you tried restructuring it in a different way, using keys, etc, I'd give it a day before the bot makers figure out how to get around it.
Very true, and I go on to show this. It is supposed to be an illustration of how this could not work as a solution.
Perhaps I did not make this very clear, sorry! I think I will add a clarifying line for this.
Damm so long post , this days im intrested about bots . Im not a fan of this . Any voting bot account must be consider and count = 1 vote even if ads 200 votes in totall . This will make clear how many people read or look at posts . About analistik boths its ok i found the idea very good .
Thanks for reading! It probably is a bit too long
I tried to make clear that "normal" accounts and bot accounts are treated exactly equally by the Steem blockchain and this is a central and important feature.
I agree with you as it pertains to the block chain, but I think @patelincho is on to something w.r.t. the steemit web site. Should 100 votes at 1% really show as 100 votes, or should the UI change the display to report it in a way that's more meaningful? I'm even seeing votes submitted at a fraction of a percentage now. Maybe the web site should report ceiling ((sum (all vote percentages)) / 100) or just ignore votes under 50% or report fractional vote totals or some such thing when reporting vote counts?
I would support a UI change due Steemit.com with something showing votes in a more meaningful way, and payout for that matter, which I've argued for elsewhere.
It would be cool to see a detailed proposal for it.
That was a long post, glad I made it through it! My thing with bots is they shouldn't be up voting or down voting content. Flagging should be another option left to real people. Implementation of a captcha on items we want humans to do might be the best answer.
Yes it was, thank you for reading it!
Captcha really can't work, it could only work on a website not the blockchain. In fact this was a central point I was hoping to get across in the post ๐
We could slow them down with flagging limits per day, voting limits per day and make that the case for everyone. You could also let users vote on accounts stating if they are human or not. When enough accounts downvote their activity could be limited or stopped for periods of time. This would allow real people to police bots that behave badly.
I would not support any of these ideas, but it's interesting to see them none the less. Especially determining on an official blockchain level whether an account is a bot account or not, it should not matter.
Under my Hard Fork Solution proposal though, you'd be able to down vote posts by suspected bot accounts if you disagree with bots and do it that way. Though I wouldn't personally agree with that I think it would be a totally reasonable way to use your voting power.
I think requiring on-going activity to enable voting would be harmful to the platform in the long term.
I have used theoldreader.com daily as an RSS reader since they launched and Google Reader before that, back as far as maybe 2006 or 2007. Many people there read and "Like" but don't share links. I think this type of application would be a perfect use case for the steem block chain (in fact, I e-mailed the support team at theoldreader a couple months ago to suggest that they might want to make use of steem - though I never heard back from them). Also, many people simply prefer not to comment or post, but would still want to vote.
The requirement to post regularly in order to vote would fence out people and applications like that, which would limit steem's growth potential.
The read-only bot that you discuss is basically just a recommendation engine.
Sure, it'd be fine with that, but I don't see any benefit from preventing the bot from just pressing the vote button. I should be free to vote my influence using any criteria I want. Maybe my bot correlates certain keywords with a rising steem price and decides to vote for those. It doesn't need to read the article to decide that the post brings value. Why should it have to wake me up in the middle of the night to press the upvote button for it?
Up-vote doesn't mean I like everything about an article. It means I like something about an article enough to want to say so. Bots have an advantage because they are more productive, and with the right incentives in place, they'll eventually be better than people at surfacing good content. Instead of trying to hobble and discourage them, we should encourage them to improve, which is why I am totally on board with the continuing need for competition among bots.
I think you are correct about restricting the usage of Steem, it certainly limits the applications. At this stage it could be too much, we don't know the full potential of Steem yet.
However the restriction may not be too large on the kind of applications that use Steem as a social network first and foremost. Don't forget that this was the original and express intention of the creators.
Of course the read-only bot is a recommendation engine, or a filter as I prefer to call it. To be clear I'm not saying bot voting should be disallowed, I go to lengths to show it cannot be done without corrupting Steem. Rather that some bots may not offer auto-vote features but the rest of the stuff that usually goes with them, i.e. the filters.
Glad you agree with bot competition, rare to see that stated ๐
I read about a construction design method that was used at some college a while back. Rather than lay down sidewalks between the new dorms and classes, they left everything unpaved. Then, when the students wore the paths into the ground, that''s where they laid the walkways.
Here at steemit, I see soo much effort in the opposite direction. Every time someone finds a way to succeed, instead of emulating it or competing with it, an overwhelming chorus of voices emerges cheering for the avenue to be blocked. I know you didn't say that bot voting should be disallowed, but I suspect that's what many people are thinking when they read about a read-only bot. And you sort-of did imply that bot voting should be stigmatized (which surprises me, given fossbot).
I feel like this post was much less clear than I was intending ๐ญ
I recognize that bots are a perceived problem by many. All I intend to do is take their concerns seriously andโ explore certain realities and potential realities, showing why they must be allowed for and teasing out the impact if they were to be prohibited.
Indeed, I'm in favor of them, if my position is in any way unclear. I will reread this post and make a few clarifying edits, you're not the only one!
By the way, absolutely beautiful metaphor for experimentally lead directioning. You should do a post about that, it's powerful.
I don't know what you're referring to exactly when you say Steem tends to go in the opposite direction. Do you have any examples?
I'm probably reacting to the comments and the general sentiment towards bots as much as to what you wrote, so maybe I read too much between the lines. If so, I am sorry for that.
I see that you joined in December. Maybe the other dust-ups I'm thinking of happened before that, but there has been high drama a couple times when people managed to find innovative ways to consistently get to the top of the trending page. Specifically, @steemsports, might have blossomed into a full-fledged prediction market if they hadn't been driven away to their own web site (or maybe they still will, I lost track of them when they left.), and @steemvoter was also bullied into near silence by a mob for trending too often.
And lastly, of course I think bots will be one of steem's best features, but there has been an informal anti-bot campaign on the site since I joined in July.
I hope your approach builds some bridges on this topic. Maybe you already saw my on bots series, but I'll mention it in case it contains any helpful thoughts.
I am not a vintage user, it's true ๐ I've definitely missed some important things. I was around for some of the objection to @steemsports. I actually really like what happened there, people objected and had the power to make their objections felt. Sounds ideal.
I was also there for the @steemvoter objections and personally supported those objections.
I thought you were referring to the "management" moves in the opposite direction. If the users on the platform object, that their objections have some consequence is good in my opinion.
It's the "crabs in a bucket" thing playing out as the platform designers intended. From the whitepaper:
I also think that bots could become one of the best things on Steem. We just need more variety and importantly, more of them open source! I intend to build bridges so to speak, thanks for this note.
I did not see your series, I missed it! That's one of the main reasons I created the fossbot, to find stuff like this. Followed and will read these now.