Downvoting has gotten a lot of attention recently, with even @dantheman writing 2 articles about it.
https://steemit.com/steem/@dantheman/negative-voting-and-steem
https://steemit.com/politics/@dantheman/the-politics-of-negative-voting

It's important to read both, they provide context and this post is meant as an extension from this background.

Mostly though I want to focus on this idea, from @dantheman:

The goal of Steem is to reward users proportional to the value they bring. An objectively bad behavior is one where a user manages to get large rewards while providing others with little value.

There are really two goals here:

1. Reward users proportional to the value they bring
2. Prevent large rewards for low/no value

Later @dantheman discusses the issues with gaming the system for whales, so I’m going to add:

1. Prevent powerful users from gaming the system to their own benefit.

Hypothetical situation

Let's assume there are two imaginary groups: the pro-sentient plants group, and the anti-sentient plants group. Let's call them the Pros and the Negs.

These two groups both begin providing value, writing good articles, getting upvoted, and beginning to build following. This is good, because this increases the value of the Steemit network.

And then they get pissy

Somehow, the members of these two groups get upset and begin to downvote each other's posts into oblivion. This begins to hide such posts on the site, destroy reputations, sap both groups earnings, and makes Steemit a hostile place in general. In the long term, Steemit is worse off.

If one of these groups is substantially larger than the other, then it can overwhelm the smaller group and essentially censor them from the site. As the downvotes add up and reputation drops, the smaller group has less and less ability to fight back. In addition, if the smaller group tries to make peace and DOESN'T retaliate there is no way for them to save themselves the damage, other than to bring in more people to upvote (harder than it seems).

Now, what if the larger, hostile, group begins downvoting these allies? Unless those allies have tremendous Steempower, they will quickly find themselves overwhelmed.

A NOT so hypothetical situation

China, Russia, and other nations have well funded social media organizations that have vast resources. In this case, they may actually want something like Steemit to succeed because unlike other social media platforms, you can BUY influence. For a nation-state, that's a godsend. They have the time and resources to create a ton of well-funded sockpuppet accounts. These accounts wouldn't attack the system as a whole, but would instead devote that influence to targeted censorship: pushing certain stories to "trending" and silencing opposition.

These actors could invest tremendous resources into Steemit but not be "invested" in Steemit

Investing and "being invested" are different. Throwing some money and time at a problem is not "being invested" for a nationstate or other similar organization (corporations could do the same). It’s not the same mentality as an artist posting day after day trying to build a following. For an existing, powerful person or organization it's just a strategy, a disposable tool. These organizations don't care about the success of the whole, just ensuring that the site can be used as a tool to achieve their ends. Right now buying in for $40k would be sufficient to wield tremendous power. Spending 10x this is just a single line-item for such actors.

Downvoting as it exists is a threat to the Steemit ecosystem

Let's look at the ways that downvoting can be hostile to the network:

1. Groups downvote each other and eventually one or both no longer wishes to remain on the site - bad for Steemit. Ideal setup would be both groups able to produce value but unable to affect one another.
2. Large, monied organizations have both the time and resources to buy/create whale-like influence and use it to target subjugation on subject matter they don't like.
3. Censorship of any kinda is hostile to a blockchain based network that is opposed to censorship intrinsically.

What would a solution look like?

Let's talk about requirements before I propose my solutions. What we need is the following:

1. A system that allows both up and downvotes for curation purposes
2. A system that keeps groups that "can't play nice together" separated and able to produce value in peace.
3. A system that restricts the power of potentially hostile actors from using their power on the network to censor views they don't agree with.
4. Doesn't rely on whales or otherwise "riding to the rescue". That invites tremendous politics and politics of this nature tend to create distrust in the system and damage it.

All 4 are important. Especially with the way rewards are distributed, it's inevitable that there will be a lot of concentrated power in a few hands. This is itself worth of another post because that has other issues outside of downvoting (but similar patterns can emerge).

Proposal 1: Blocking of certain users

Allow users to "block" other users. This blocking would have a few key effects:

1. It would "mute" the blocked account.
2. It would remove the ability for the blocked account to make any status/payout related actions for GOOD OR BAD. Their upvotes and downvotes would be completely ignored.

These two aspects would help keep groups separated. It also discourages people who "block" people for a single or justified downvote because they would lose the potential payouts of that same person in the future.

Proposal 2: Force comment with downvote

If someone's content is bad, or you take issue with it, or whatever, the only way that user can improve is to get feedback. A downvote doesn't provide context, it doesn't provide a why. An upvote doesn't either, but it's a lot easier to infer: "I think this is good content for the network". An upvote says "more of this". But a downvote could be about an aspect, like bad writing. What does the author need to fix to make it up-voteable?

Proposal 3: End self up-voting

@dantheman notices the issue with self-voting in his own writing when he talks about the "upvote-only" economy. Here a whale could just block all other whales and vote for themselves over and over. It doesn't stop whales from having proxy accounts to do the same thing, but the hassle is greater. Fixing the proxy account problem is an isuse for another day.

Ending self voting also just makes sense: If you don't think your post is great, why did you write it? The shear production of a post and the work involved is an intrinsic up-vote since you only can make so many posts a day before things get diluted.

This still doesn't solve one of the REAL underlying problem he identifies which is the need for other whales to act as payout police. Having payout police like this will not scale as the user base gets too big, or even too-multilingual. In the case of current whales, @berniesanders has been working to try and balance payouts via downvoting but that too has lead to worries. His balancing of @dollarvigilante lead to a lot of controversy. Single human judgements are prone to that. In addition, I doubt the current whales can read/write every language or even have the time/energy to try and track all this. Future whales may very well be the corporate/nation sized actors mentioned above which can definitely not be trusted to be effective monitors on the network.

The second remaining issue is what to do about whales that just create proxy accounts and use their high SP account to upvote everything they do (to get around the upvote limitation).

Proposal 4: Fast die-off for voting power on same person

This is to help mitigate the proxy problem. If account A upvotes the content of Account B, their power to upvote B is reduced across other posts. This penalty would be cumulative, and the re-charge would need to be long. This isn't a perfect fix, but it makes everything more complicated. Whales would need many accounts, would need to use all their votes across accounts, and then would not get an easy transfer. Can this still get gamed? Yes, but it's harder and the behavior patterns of gaming would be very easy to detect as well and then subsequent approaches could be taken.

Proposal 5: Allow curation tags

Allow curators to “tag” posts in ways that are not used for organization but just for metadata. For instance, curators could flag content in informative and evaluative was such as “badly written”, “chinese”, “thorough”, “dishonest”, “scam”, “adult”, etc. These wouldn’t be used for reputation or anything (yet?) until we began to see how it played out, but would help enrich the ecosystem.

Remaining issues

• How to handle powerful bad actors from forming a hostile cabal against the network for external gain?
• How to deal with the follower problem: just upvoting popular users and getting free payout, regardless of quality. This strategy gets worse the more people do it, breaking the point of curation.