Challenges

This is a long and incomplete list of challenges which needs to be edited, completed and prioritised (21 May 2020)

The ambition behind offerbots is enormous – to bypass aggregators by providing everyone with a personal tool for making and receiving offers.

The challenges that this goal presents, however, are legion. Here is an incomplete list of issues which need to be addressed.

Diversity

The offerbot diversity paradox

Offerbots are an attempt to radically increase the diversity of our economy, society and information by creating a uniform (federated) approach to offer processing. This uniformity will naturally limit diversity if we don’t identify ways to limit or reduce that uniformity as we build traction.

The offerbot opportunity paradox

Offerbots are intended to greatly increase the number of opportunities available to people by bypassing aggregators with a free tool. If successful, this replacement will result in the loss of opportunities to in the aggregator ecosystem, particularly employment with those aggregators.

Diversity versus thickness

One of the desirable attributes of markets is that they are thick, meaning that they “bring together a large enough proportion of potential buyers and sellers to produce satisfactory outcomes for both sides of a transaction” (Al Roth).

Aggregators foster thickness and increase their profits through homogeneity. The Yellow Pages, for instance, had broad categories for services such that many service providers were listed in each category (and they had incentive to pay the directory to stand out from all of the other service providers).

The only way that offerbots can provide thickness and diversity is to have a large number of people using offerbots (such that each small category of products, services and information) is populated. Even then, the diversity of methods to create markets (e.g. different concepts for the word ‘bicycle’) might fragment markets to the extent that they never become thick.

Diversity versus feedback loops

Homogeneous systems benefit from strong feedback loops. The reason that we send an SMS, for instance, is that it our message is reliably delivered to and seen by the recipient (even if the phone is on silent or switched off, it will be seen). This is also the reason that we might not SMS someone if our message is not urgent – we might send an email instead.

When each person has control over their own offerbot, there is no guarantee that their offerbot will function as we expect. Someone might send a message in an offer and mark it urgent, but the receiving party might not recognize the urgent tag (or it might function them differently).

This diversity will limit the strength of feedback loops and may slow the adoption of offerbots. There will be solutions – e.g. sending receipts to confirm the owner of the offerbot has viewed an offer – but each of those solutions also require their own feedback loops.

Complexity

Execution complexity

Starting an aggregator for a single market is incredibly difficult, mostly due to the chicken and egg problem in two-sided marketplaces.

Transitioning the world to billions of new offerbots-based markets (without vast sums of money and while competing with aggregators who do have vast sums of money) will require world-class execution.

Networking

Distributed Denial of Service (DDoS)

Normal levels of traffic in a distributed system might have all of the challenges of a DDoS attack on a centralised system. On top of that, we need to deal with malicious DDoS attacks.

Priority

Owners need instantaneous, prioritised access to their offerbots in order to serve data and views to their apps and browser.

Limitations

Diversity versus Attention

Economic diversity comes from distributing our decision-making:

The problem is that our attention is scarce and, therefore, we have limited ability to Instead of everyone using the same small set of vendors recommended by Google Search (say),

Imbalanced processing power

The affordability (or lack thereof) of processing power will result in at least two problems.

All offerbots will be peers, but some will have access to orders of magnitude greater processing power.

This is not necessarily a bad thing – these larger offerbots will act as supernodes in the system and contribute to the decentralized layer for offerbots (which are otherwise distributed).

Missing Infrastructure

Payments Infrastructure

Our payments infrastructure does not support micropayments. A currency created by Facebook is almost certainly not the answer.

Privacy & Security

Privacy versus Resources

Your preferences may be disguised from the owners of other offerbots by requesting broad sets of offers and filtering the large number of offers received locally.

People with little processing power, bandwidth or storage for their offerbot will have incentive to reveal their private preferences (by requesting a narrow set of offers from other offerbots to the costs of filtering them locally).

One solution for this might be the use of proxies, including onion routing.

Distributed Security

Aggregators offer a feudal model of security. That is, they take care of data security for us but at the cost of our privacy and the distortion of everything we care about.

Legal Intercept

An increasing number of governments have laws requiring that they can intercept. Other governments cannot be trusted with private data.

Illegal and immoral offers

Every communication technology is used to offer illegal and immoral goods and services, and offerbots will be no different. The question is whether we can do anything to limit the bad without breaking everything else.

Family controls and self-imposed limits

Children need protection from adult content. Some adults would like filters to limit their impulsive behaviors (e.g. gambling or porn).

Hosts

Hosts must not access user data

If hosting offerbots will give hosts access to private data, bad actors will host offerbots and steal private data. Offerbots must not, therefore, give hosts access to that data.

Hosts may run malicious code

Hosts may claim to run trustworthy code but replace it with their own.

Externalities

Carbon

Centralized data centers have an enormous carbon footprint. Distributed hosting will be even worse.

Jack of all trades, master of none

Aggregators have enormous resources for optimizing their user experiences and

Distortions

Undeclared distortions

Some celebrities recommend products without declaring that their endorsement was purchased. Some providers of offerbots will favour and/or exclude vendors from markets which they claim are neutral.

Identity

Verifying identity

We will need to assess the trustworthiness of third party verification systems.

[To be written up]

[This items are not less important, I merely haven’t written them yet and then prioritized them in the document … ]

Trusting unknown parties

Costly signals

Reputation

Networks of trust

Reputation systems are broken.

Privacy

We currently choose to give away our privacy because we need external offer processing (and the only form of external offer processing made available by aggregators is one in which we must surrender our privacy).

That said, there are still privacy concerns with offerbots.

Firstly, privacy requires that we do not reveal our preferences to other parties which, in turn, has a real cost in terms of processing power, bandwidth and storage. Consider, for instance, that we need to buy a specific medication from a pharmacy. We could send our offerbot to ask their offerbot whether they have that specific item in stock, but doing so reveals something about us (that we, or someone close to us, needs that medication). The alternative is to obfuscate our request by sending our offerbot to asking for all or many of their offers, but this costs processing power, bandwidth and storage, all of which cost real money. People with less money, therefore, will have incentive to reduce their levels of privacy.

Secondly, there can probably be no implicit or enforceable agreement between parties who own offerbots. Certainly there is the law, but it would be difficult to identify cases of offerbot owners selling data about the offers requested from other parties. Worse, it’s entirely possible that the makers of an offerbot might do this without the knowledge of the owner.

There is a clear trade-off between processing power and privacy in that

  • privacy versus processing power (asking for something precise saves processing power but reveals your demand)
  • privacy in asking for embarrassing things from the wrong people (e.g. HIV meds from work colleagues’ OfferBots, which may log your request)
  • privacy onion routing
  • privacy – agreements, third parties, etc.
  • Energy usage (distributed rather than centralised)
  •  Regulations – Lawful intercept versus privacy, content regulations (e.g. removal of violent videos, copyright, etc.), protecting individuals from hosts
  •  world-class execution – solving the chicken and egg in every market.
  • feedback loops – strong feedback loops drive participation but homogenize activity. How can we facilitate many (weaker) feedback loops?
  • diverse markets may not be thick (not many participants).
  • identity verification and trustworthiness
  • safety without centralised curation, filtering, boundary-setting, etc.