In the previous article, we looked into the key purported innovation of the Algorand project, namely, its pure proof-of-stake consensus mechanism. Based on a version of Byzantine-fault-tolerant algorithm and potential participation of every tokenholder, it ostensibly allows to add blocks to the chain at the speed of their propagation and guarantee their genuine finality, unless more than ⅓ of the tokenholders are malicious. If it really works as designed, it could be a true gamechanger in the public blockchain space.
Consensus without incentives
However, Algorand’s approach has not been without its critics. The most important reproach that they made concerns the absence of any attempts at incentivizing the would-be consensus participants to act honestly or to actively participate. There are even no block rewards distributed to the proposer of each new block or the 1000 members of the validating committee.
Researchers from the Blockchain at Berkeley initiative have raised several important more precise concerns about the incentiveless consensus scheme. Arguably the most pressing one concerns liveness or the guarantee that the network will not experience halts in transaction validation. The reason to worry here is that in order to participate in Algorand’s consensus procedure, tokenholders must continuously expose their private keys. This puts them at risk of hacks but this risk is not compensated by any rewards. Berkeley experts believe that this may cause a lot of tokenholders to opt out of participating in block validation and hamper liveness.
Secondly, the lack of rewards for block validation means that those who buy Algorand tokens will have to forego other opportunities to put their funds to work without any direct compensation. This implies that there will be little motivation to acquire Algorand tokens specifically for validation. The Berkeley researchers do not develop their argument beyond this but it arguably reinforces the previous concern. If users who hold Algorand tokens for other purposes do not participate in validation for fear of exposing their private keys and there are no users holding them specifically for validation, then who will Algorand’s consensus mechanism rely on?
To his credit, Algorand’s mastermind Silvio Micali has not shied away from responding to the lack of incentives criticism. His viewpoint can be summarized as “incentives are the hardest element to get right, and it is better to have no incentives than bad ones.” As an example of the latter, he prefers pointing at how the creator of Bitcoin Satoshi Nakamoto had never imagined that the system of incentives he envisaged would lead to the emergence of massive mining pools that dominate the mining market and may engage in reordering transactions in blocks, for instance.
However, while Micali’s retort may be valid as far as it goes, it does not adequately address the practical concerns about Algorand’s approach. This means that we will probably need to wait until the launch of the mainnet to see whether the lack of incentives will turn out to be truly damaging in practice.
A problem in a way similar to the lack of incentives with regard to consensus may concern Algorand governance that is supposed to be conducted in the same way as adding new blocks to the chain – through Byzantine agreement by randomly selected committees of 1000 validators. The only major difference is that, instead of a block, the governance votes may concern monetary policy rules, for instance.
Projects like EOS, Tezos and MakerDAO have arguably already demonstrated that blockchain votes almost invariably suffer from low participation, especially by smaller tokenholders. Crypto thought leaders such as Vitalik Buterin (for example, here and here) and Vlad Zamfir have in the past provided insightful commentary on the potential consequences of this, including outsized influence by whales, low perceived legitimacy of supposedly binding decisions, the potential for vote buying and external attacks on the network by actors owning only a small percentage of all tokens.
In addition to the concerns related to low participation, even if it were possible to induce a significant number of tokenholders to engage, smaller tokenholders would still have few incentives to approach their votes with diligence and seriously consider the alternatives, as well as resist even small bribes. The reason for this that the probability for a small tokenholder to decide the outcome of a vote would be minuscule.
Algorand supporters could respond that the absence of privileged nodes discussed in the previous article, as well the intended approach to initial token distribution using a Dutch auction without any pre-sale or on-chain transaction fee races (like in BAT’s case), where the market will be allowed to discover the fair price, will result in a much less pronounced concentration of token ownership than is usually the case with native blockchain tokens. Which will, in turn, mitigate some of the problems of outsized influence and vote buying.
One has to note, though, that this solution did not achieve the decentralization enhancement goal for the Gnosis community that tried it back in 2017. Also, while the initial wide ownership distribution may be achievable, it does not mean that it will not change over time. Nor does distributed ownership somehow in itself make tokenholders more diligent about their votes.
Nonetheless, it is difficult to tell in advance how important governance issues may become for Algorand in practice. So far, governance has arguably been messy at best for all the serious public blockchain projects, yet none of them has so far been doomed by its shortcomings. It would thus seem that, as with consensus issues, only time will tell.