Public Decision Making - Letting Billions Create the Future Together
Next: Examples of Public Service Platforms
Decision Making
In previous posts we’ve discussed old and new decision-making models like representation, conviction voting, quadratic voting, prediction markets (latest ones in the segment about decision making in decentralised organisations) etc.
One characteristic of public issues is that there are exceptionally many people that are affected who at least indirectly act as decision makers (aka vote for representatives). You cannot opt-out of legislation except by voting with your feet.
The amount of affected parties on global level is staggering. Would it be possible to have more direct models and how would they look?
How a million or billion persons could make decisions together?
When solving global problems, everyone should be involved. Everyone can have insights.
Everyone who has worked in a large company, knows that when different proposals are presented for decision makers, already perhaps 90% of the decision is done. Huge number of alternatives have been ruled out. Also, alternate investment concepts that the funds could have been allocated and that may have been much more useful and profitable are not there (this is called opportunity cost).
Same is true of political decisions – they are today made in a representative system where fairly small set of folks participate into the process of drawing up alternatives. The way of working has developed during times when communication required travel and travel was slow and expensive.
These constraints no longer hold, and decision processes can also be re-thought and the right models found through experimentation. This does not necessarily mean that representation is a bad concept but with modern tools it can be improved.
Another issue is that as populations have grown, less and less people have the opportunity to personally know politician. Most information comes through some digital means. Increasingly via social media. As there is competition for attention, this favours short, eye-catching messages with clear division between what we propose (good) and what is bad. Politics becomes more and more like advertising.
How could decisions be made with people by people and for the people by involving everyone or at least a very large group of people?
Things start when people gather to discuss and share views and opinions and getting feedback and improvement ideas from each other. Discussions between people need to happen in relatively small circles as only one person at a time can be speaking at any given moment. How could one billion persons discuss in parallel some topic in a meaningful way?
One approach is for everyone to divide into small groups (max 8) that are physically in the same place. The basic idea is that people first talk with each other for some time on the topic and write down or use voice recognition to capture ideas.
Computers are good at taking huge input and clustering and analysing it to a format where common ideas are put together and showing linkages between concepts. So, in next step computers would take all this input from a billion people and try to categorise it into meaningful representation.
The original teams can then see what the collective hive mind has come up to get some new perspectives. Pick up things they particularly like, start working on them and make comments and suggestions to make them even better.
Rinse and repeat until some patterns start emerging – what type of approaches gain most support?
The top alternatives could then put through the analysis and simulation phase to check for validity. Then field trials and finally into decision making. One or more options are then going live and monitoring starts (to check whether they actually work and do not just sound good and work well in simulators).
Or proposal could go via Open Ministry or Citizens’ Automation Budget type of models also. Latter one could implement/trial several proposals in parallel.
One could call this augmented decision making because computers are there to help out, sort and shift ideas to make them easier to understand so that large body of people can participate.
This is not the only approach for involving large groups in collaborative proposal forming (remember the 90% rule above). I’m also sure the gentle reader has heard of the general rule of any statistics or percentage quoted: “47% of all statistics are made on the spot, including this one”.
Let’s discuss self-organising “Pyramid” (do not know a name for this model, if it has one that is commonly used).
Idea is to have a hierarchical setup where each small group of roughly eight selects one of them as representative to the second layer. Members discuss and form their own view and their representative takes it to the second level that consists of selected members of other small teams (about eight again).
Second layer discusses, forms an opinion, takes back to first level to get comments. At some point a consensus start forming. The second layer also selects a representative to a third layer and so on until there is a small set of representatives at the top to effectively negotiate with each other.
If the group size is 8 and there would be 200 representatives in the highest level, 7 layers would be needed to represent 52 million voters.
Once a decision is made, it is either published and/or it is the responsibility of their representative to explain to previous level what was decided, why it was made, what the simulation and experimentation results were and how the situation is monitored if it is successfully implemented. This explanation is down in a downward fashion at all levels. If teams on any level are not happy with their representative, they can change this person.
Those are two options for letting million or billions to work together in self-organising manner.
Let’s turn the attention to other topics in decision making that need some airing.
Training Decision Makers
To improve decision making, there are lots of different frameworks that go under names such as “how to solve complex problems”. The core aim of these frameworks is to help people who do not share common world view to find out approaches that they can agree on from different viewpoint (i.e., what common stuff works for all). The very aim is not for people to have a common view on the topic but to disagree. From own viewpoint people then find solutions that all think are acceptable.
They typically follow a set of steps where you identify affected or participating people (stakeholders), what factors affect the outcome and how they are linked together and help people get ideas how to modify or affect those factors.
It seems natural that such training should be part of school curricula and for older generations online training and peer mentoring. It’s not.
Computer Aided Opinion Forming
How should the computer programs organise the crowdsourced materials? From Taiwan comes one alternative. They have been using on open-source solution called pol.is that aims to find consensus among a myriad of views on a topic.
Polis lets anyone share their feelings on a topic and agree or disagree with what others are saying. But instead of trying to maximise engagement like other social media platforms, it draws a map of agreements and disagreements as they emerge and it promotes agreements that reach the largest audience, not just people inside own bubble. At the same time, it depromotes trolling and provocative statements. As one example how this works, in a heated debate on Uber driver regulation, a consensus emerged that people just care about safety.
For more see:
Language
Above are ways of involving people in the process before decisions are made and ways to help people as they go about doing it. But there is a catch and that is to do with language.
Today the people participating in such assisted decision making do not even have to speak the same language as automated translation services are increasing common. The decision process can surpass national borders and participants could come from anywhere in the world. For example, development aid decisions could be made together between folks in the receiving country and the tax payers in the donating country through such a direct and iterative process.
There is however a problem with language. People think that language is a common tool for interaction and co-operation between people, but this is not so. Since the experiences and lives lived differ, also the exact meaning of words varies between people. Different languages split the reality into different size “buckets”. Language is affected by the environment surrounding and paths taken by speakers in the past. Language is born and shaped in human groups all the time. In other words, people think they share the meaning of words, but there can be great variations that people cannot see.
This is greatly amplified in global world, where cultures and identities differ greatly. There are words and concepts practically in many languages that do not have direct equivalents in any other language. Words also commonly have multiple meanings (called homonyms) and the meanings are only partially overlapping between two languages. Sometimes a word means something and its opposite (like word scan in English to look at something broadly and in detail) Some people like to have fun with such fuzziness of language by saying something and its exact opposite in the same sentence. Even simple combinations can have differences. White house in US can have different meaning than white house anywhere else.
This is an area where machine learning could be applied. To at least warn folks communicating that the other person has different overtones in their sentence. Perhaps in future even outlining what they mean and why is that.
There are more subtle questions in translation that are hard to solve. Should translations try to express the meaning and radically change the structure and words from the origin, or should systems translate the words as accurately as possible or should it try to preserve the structure of the expressions? The desired method probably depends greatly on the context. Cultures have preferences that differ. Some tend to appreciate more poetic ways of saying something, others value greatly directness and sharpness. Should translation respect the habits of the source or destination language?
This is a type of a question to which there are no real answers. Different human translators have each their own style and local publisher select one that they think suits the book best. Could there be multiple translators available so that people can pick the one that works for them best?
Credits: Core concept is presented in Timo Honkela’s (not translated) book ‘Rauhankone’.
Another fundamental question is that to what degree language shapes or limits the speakers’ worldview. This idea is called linguistic relatively or Sapir-Whorf hypethesis. It is no longer believed to be true but a weaker version where language's structures influence and shape a speaker's perceptions, without strictly limiting or obstructing them has empirical evidence.
For deeper level understanding, it may still be good to self learn other languages rather than rely only on computers. Computers naturally can expand the scope to - in principle - all spoken languages.
Side Note: If computers ever start to understand human communications well on the meaning level, they can also be used to detect manipulative patterns in speech and warn us.
These are always based on our cognitive biases. For example, if something is liked by a lot of people, it most likely is a good product. An indirect way of knowing that something is popular, is its scarcity. If something became just available and is already sold out, we think that it must be good. Or reciprocity, if I get something for free, I feel obliged to somehow pay it back, otherwise I seem like a selfish person.
Computers could detect these manipulative plays and warn us in real time and suggest countering responses.
That’s all for this week. Next week we’ll start a mini-series on examples of new types of public services.