Open Ministry/Municipality
Next: Governmental Kanban
In previous post set we discussed Citizens’/Enterprise Automation Budget concept of making public side more directly responsive. It’s a way of making existing services better, but does not allow citizens to contribute to the evolution of the public side.
Open Ministry/Municipality
Some of the common complaints about public service resolve around the times it takes governmental or municipal authorities to take action when a new need emerges. It takes years before legislators notice when a new type of scams is invented and even longer on grey-area mechanisms, but people start getting burned immediately.
Why not crowdsource ideas for legislation and give citizens a say in what areas of government should focus (i.e., to prioritise areas of interest)?
The concept behind Open Ministry is that any citizen can make legislative proposals and if enough people support it, the proposal needs to be formally handled in the parliament.
This is a mechanism, where input for new laws and changes comes directly from citizens. Proposals do not have to be final and well written, but idea must be clear. Ideally this is followed by an open discussion to improve them. Once the initiator thinks it is ready, it goes for voting. If it gathers enough support, it needs to be handled by the parliament in official process.
Open Ministry concept is currently in place in Finland. Current implementation does not include the stage for working and improving proposals, for some reasons politicians have not seen this as essential…
If my memory servers right, it started as a private initiative by a single guy who had own money for a thousand euros and did not know how to code. Somehow we managed to implement a web site (now defunct) with ruby on rails that allowed crowdsourcing legislation. And so he ended up changing Finland and perhaps more parts of the world for the better. (and for the world, cannot find any links to that person for checking this).
There are many ways how this crowdsourcing concept can be improved.
To make relevant proposals, people need visibility. This is today lacking. Visibility as to what is the state of affairs now (i.e. open API for various statistics and details ( more on that later) and what type of changes are currently progressing.
States and cities need an enterprise planning system (ERP) where residents see how issues are progressing in the administration and decision making. You filter for example for zoning and see what is at idea stage, what is being discussed at board level, almost ready proposals where statements from stakeholders are called for and what is ready for decision making.
One question for future is naturally that why does the proposal concept be reserved for citizens of that nation only? Why wouldn’t it be allowed say for someone from India to make proposals to Finnish legislation? The reason for this is that we want to get the best ideas onboard. For that Indian person the reward is both recognition and since if they truly believe in what they propose, this way they can get it implemented at least in one country as a proof point and use internally in their own country as an argument.
Overview of how open ministry/municipality could work?
What type of flow would ensure that the implemented changes work well?
Analysis
The simplest analysis is to see if the proposal is already a law. In general people are not aware of laws in place and regularly propose something that is in effect already. More common than you think.
Interference analysis can be done either manually or automatically although today laws are encrypted in a hard-to-understand format for both machines and people. If there is interference between laws, rules for precedence needs to be proposed. As the interference can be complex this may make simple precedence rules impossible in some cases. But at least connections are detected enabling discussions about them.
Governments overall collect huge amounts of data for own planning and for statistical purposes. Today simulations are done for the most important laws, but the models used are not made public, nor is there explanations why the model is built in specific ways, whether any study was done how sensitive the model is to small changes in starting assumptions and so on. There is some weak evidence that the models are often complex looking spreadsheet that look “sciency” but lack any attempt to consider for example dynamic effects (dynamic effect => one change affecting multiple other parts of society that in turn cause ripple effects in many places). And we know that at least in corporate world 80% of spreadsheets contain some kind of calculation mistake (fear of this may be the reason why public side wants to keep them secret).
Nor do citizens have any access to the big data collected except as highly aggregate statistics. When data processing was manual, this made perfect sense, but today there is no technical reason for this.
The collected base data can be made available in a sandbox environment via an API so that anyone can access the data and make their own analysis. It improves transparency and general understanding of society if the existing official analysis models are open sourced and made available on a modern lab-notebook style tools (think Jupyter notebooks, if you are machine learning bent or have exposure to those methods). In these tools you can see every step how the model was built with explanations and the final steps used to run the final results. Using the same notebook anyone can re-run the analysis, make small changes to test if some parameters were selected under political guidance.
Simulations are another way to approach analysis. Computers allow to run impartial simulations without any preconceived ideas of human behaviour. Just simulating what are all the possible outcomes from a given starting position with the rules in place. Are there any corner cases that produce really bad outcomes and what those would be?
This in contrast to current models of economic theories that take in quite strong assumptions about human motivations – i.e., assuming logical beings withe near perfect understanding of current affairs maximising own financial success as the only valid motivator. This makes economic models simple and enables nice careers for people who study and teach them, but never shall they meet with the real world.
Impartial simulations are especially important because a reasonable expectation for future will be that adversaries and criminal organisations will use automated simulation tools to discover loopholes in legal frameworks. Then they will build automations to quickly benefit from them. It would be foolish for the public sector to ignore this.
Simulations uncover faults in early proposals and allow to iteratively improve them.
Simulations enable among other skin-in-the-game analysis. Are the proposal makers expecting someone else to bear the cost while they benefit for the change?
In sensitivity analysis small systematic changes are made to the proposed model and assumed starting conditions to see if miniscule differences would swing the results to a widely different direction. This is not as uncommon as one would like to believe. Advertisers maximise the benefits of their offer by carefully selecting starting and stopping dates (common in investment promotions or selling heat pumps as examples). Also in a highly complex and interdependent world there almost always are unintended consequences that may be small or big. The purpose of the sensitivity analysis is to see how robust the proposal is for changes and how much the proposal makers tainted their initial figures.
Automatic analysis can also be made of cost implications – would it reduce or add costs? If it adds costs, what does the proposal maker suggest, where would funding come from – higher taxes, removal of existing governmental functions, increasing loan taking or is the funding unclear?
It is today standard practice that all stakeholders are offered to state their opinion on the proposal and what arguments they see for and against it. The above mentioned simulations and sensitivity analysis can be a big help in this.
Refinement
Ideas are open for comments and proposal makers will refine them based on analysis. Once the proposer makers think they have reached maturity the ideas go for go/no decision on field trial.
Small Scale Trials First
In a complex world where everything is interconnected, it is almost always impossible to predict final outcome. Small scale trials tell whether the simulations are any good out in the real world. Ideally multiple variants or competing ideas are tried at the same time.
One built in feature in trials is that it is known that just monitoring tends to improve human behaviour (i.e., knowing that I am taking part in a trial). Trials may show too good results, but that is hard to prevent.
Field test will sometimes uncover that the proposal does not work as wanted. In that case there are two alternatives – either to go back to drawing board and make changes or to scrap it. People did not work as the theorists wanted. It is not possible to change people, better to adapt theories.
Reference Implementation
During field tests, an initial reference implementation helps greatly. This means there should be an API that applications can tap into and rudimentary user interfaces and automations built. The implementation does not need to be completed, an API can trigger for example manual work if changes to underlying public systems would be costly and take time. Such minimum viable implementations allow to estimate with much greater accuracy costs for full implementation.
If there is requirement that enterprises or citizens need (or are allowed to) do some actions, a reference implementation needs to be taken to a level where it is easy to take that functionality into use. For example, sample code that users can drop into their existing apps to take into use the new concept.
As an example, when legislation regarding personal privacy is defined (GPDR) as part of the legislation a reference implementation could be written that shows how a web site owner can ensure that the intent of the law is met. This reference implementation could then be used by various software tool vendors as a guideline when they ensure that their products meets the legal requirements.
Lack of reference implementation is an indication that legislators do not know in detail what they are proposing for. Unclear laws create fear and prevent economic and other human life. It is better to figure out what you want before passing that out as a law.
Over time a larger region consisting of several nations can develop common interface standards. This would mean that such reference implementations can directly be used in all member states. They could also be written as a set of smart contracts.
An alternate method is to use declarative formal definitions for regulations. Then various software generators can generate the implementation code in different programming languages and for different frameworks or smart contract environments. This addresses one problem with reference implementation – namely technical diversity.
However formal definitions have never proven very usable in the past as they increase (triple) complexity. User needs to understand both the formal definition and the details of the target system and the generator needs to be maintained when the target system evolves. The generators need constant work to keep them up to scratch. And we know that all software can contain errors, so from time to time generators fail to work properly and or the underlying system has bugs that the the generated core exposes.
One good point for formal definitions is that they allow automated analysis – if we change this part of the law, what other parts it affects and what the combined effect could be, if a new regulation is introduced, what other laws it overlaps and conflicts with?
These two approaches are naturally not mutually exclusive. One can model the regulations in a formal environment and also manually generate a reference implementation and let commercial vendors deal with additional support for different platforms and programming languages.
Monitoring
After field trials a final decision is made. Once something goes into law, the monitoring needs to continue. It’s important to see how the impact evolves and when some regulation stops working, then the whole concept can be re-evaluated.
As mentioned above, during trials people often behave better than when no observations are made. And with time someone sooner or later figures out how to misuse the system for their own benefit. If successful enough, such behaviours can start spreading. Sometimes it takes years before some pattern of use to emerge. A good proof point is to follow how Internet has changed over the years and we are still in phase where it is more or less taking its first baby steps. Same happens with all legislations, life adapts to a new environment and regulatory rot sets it in smaller or bigger quantities.
Final notes
The above setup makes the ministry quite like a web shop. Customers come in with product ideas and the web shop tries to find a suitable product and if one is not found, a set of steps are utilised to manufacture it and deliver to customers. Since the product is mandatory, everyone gets it irrespective whether they like it or not, the process is more rigorous and involves one or more democratic votes.
This naturally raises the questions whether this always needs to be so. Could part of the population opt-in to a legislation that does not apply to rest? This may not be so silly as it first sounds. States are notoriously bad at experimentation and experimentation is the root of all innovation. Perhaps cities and municipalities could be hatching grounds for all types of experiments and innovations. Residents in some city deciding that over here we like to live like this and in another city, people choose a little bit different path. Should rest of the nation mind? Maybe not always.
Caveat Emptor
Above is a rather complex system for ensuring that laws make sense. In general, making too complex proposals tend to be a good way to ensure that things never get off ground. In IT industry comprehensive approaches tend to be burdened with a disease called “feature creep” – add everything and the kitchen sink. Staring with a small scope and letting it organically grow tends to lead to better results.
Keep this in mind when reading this or any other fancy proposals.
Next: Governmental Kanban