Destructive policy

From Opasnet
Jump to: navigation, search


What are destructive policies and how should policy makers be informed about them?




Let D be a set of policies, where a policy d is defined as a selection of plausible actions that can be taken together.

Let I be a population with individuals i. The population will be affected by the actions taken.

Let v ∈ V be a value profile, i.e. a collection of value propositions about how the society should be run expressed by a group of individuals from population I.

Let S be a shared understanding, i.e. a comprehensive description of facts and values (including uncertainties and disagreements as defined by paradigm p) in a given situation held by population I.

Let p ∈ P be a paradigm of inference rules for producing interpretations from data. E.g. the scientific paradigm says that a piece of knowledge about a particular question is a collection of hypotheses that have not been invalidated using data and critical discussion.

Let k ∈ K be a knowledge crystal, i.e., a set of plausible answers to a specific factual or value-based question relevant for the shared understanding. K contains but is not limited to outcomes of interest caused or promoted by a policy d. K(d) means K given that d is chosen and implemented.

Let u ∈ U be the utility of a knowledge crystal k when valued by a value profile v under paradigm p. Also systemic and external impacts, such as climate and ecosystem impacts, are embedded in the utilities. ----arg4226: . Other, potentially important properties of utilities: Multiple minor impacts are shown by using Kantian ethics: what would the impact be if everyone did the same thing? Also, systemic disruptions are accounted for in such a way that the utilities will start falling rapidly if the risk of a systems failure increases. However, it is not yet clear how these functionalities will be implemented in practice. --Jouni (talk) 06:39, 5 November 2020 (UTC) (type: truth; paradigms: science: comment)

A policy d1 is destructive iff there exists a policy d2, d1 ≠ d2, for which the following is true for all v and p:

E(∑o U(S(K(d1),V,P))) ≤ E(∑o U(S(K(d2),V,P))),

where E is the expected value of a joint probability distribution describing the remaining uncertainties.

In effect, destructive policies are Pareto dominated policies[1] with a special emphasis on value profiles.

----arg5656: . Legitimacy should be embedded somehow. While utilities vary from a value profile to another depending on preferences, legitimacy is a social construct and dependent on time, culture, and legislation. While illegal benefits are obviously not legitimate, there are benefits that are culturally considered immoral or for some other reason not legitimate. However, one should be careful about declaring a utility to be culturally not legitimate, as often the agreement is not as clear-cut as one might think. One way to look at legitimacy is to consider whether an outcome is Pareto efficient and envy-free[2]. --Jouni (talk) 13:52, 4 November 2020 (UTC) (type: truth; paradigms: science: comment)

⇤--arg4323: . Legitimacy is not needed in this version, because now we are talking about value profiles rather than individual values. This already makes the value propositions communal. --Jouni (talk) 13:52, 4 November 2020 (UTC) (type: truth; paradigms: science: attack)

Object structures

To be able to implement this formula, each concept must be defined as objects with a specific structure capturing its key properties. This is the first attempt to define the structures.

Policy d
  • identifier and name
  • description
  • list of actions belonging to the policy
Knowledge crystal k
  • Knowledge crystal is an object that answers a specific factual or value-based question. Some knowledge crystals may have values attached to them from one or more value profiles, i.e., it has intrinsic value and is important for evaluation. An action may have intrinsic value e.g. if it is considered a sin (allow abortion) and thus has intrinsic negative value irrespective of its impacts. This may be even more important than a related impact indicator (number of abortions due to a policy). In other words, someone may oppose legal abortions even if that would increase total number of abortions (legal + illegal). Also, an actor may have intrinsic value, if it is important that a particular trusted person performs an action rather than some stranger.
  • actions are knowledge crystals answering this implicit question: what tasks are needed to implement this action?
  • actions are actual or planned work based on a policy
  • actions have the structure defined in the Kausal Watch.
  • indicators are knowledge crystals answering this implicit question: what values does this indicator get across time?
  • indicators are objects that measure an important phenomenon within a topic.
  • indicators have the structure defined in the Kausal Watch.
Individual i
  • identifier and name
  • list of value profiles that the individual supports
Value profile v
  • (value profiles are classes and may inherit properties from other value profiles)
  • identifier and name
  • description
  • list of value propositions belonging to the profile
Value proposition
  • identifier and name
  • description
  • function containing at least one knowledge crystal as parameter. The value of the function is a utility distribution across potential realisations of the parameters.
Shared understanding S
  • name and identifier
  • description of the topic described
  • written summary based on the content of the lists below
  • list of knowledge crystals belonging to the shared understanding
  • list of value propositions belonging to the shared understanding
  • list of paradigms belonging to the shared understanding
Paradigm p
----arg5656: . This and the inference rule are the least important objects if we want to keep the system simple. --Jouni (talk) 13:52, 4 November 2020 (UTC) (type: truth; paradigms: science: comment)
  • name and identifier
  • description
  • list of rules of inference used by the paradigm
Inference rule
  • name and identifier
  • description about how the rule should be implemented
Utility u
  • knowledge crystal id
  • value profile id
  • paradigm id
  • result. The result is automatically calculated for a utility based on its parameters (knowledge crystal, value profile, and paradigm) and the functions of the value propositions within the value profile. In addition to single utility u(k, v, p), it is possible to calculate a set of utilities if its parameters are sets: U(K, V, P) or U(S) where S has parameters K, V, and P. The result of U is typically summed over outcomes and then the expected value is calculated.

Value profile example

Here we illustrate what a value profile might be in practice. A discussion about face masks related to covid-19 prevention is used as input material[3]. Kialo is build on a hierarchical tree of dichotomous arguments that attack or defend another argument. While this structure is useful in many cases, here were attempt something more effective.

First, let's split the statement into coherent pieces, which are implicit in Kialo.

  • There is an action behind the argument: "The [city/country] mandates people to wear face masks in public spaces."
    • This action can be specified in several ways: recommendation or order; during rush hours; indoors only; in public transport; if 2 m distance cannot be kept,...
  • The action links to a knowledge crystal "fraction of people using mask in public spaces".
  • This links to another knowledge crystal "probability of acquiring covid-19 per trip to a public space with / without mask".
    • This already implies that there are other factors involved than just the mask, e.g. the background risk, but we do not explicate those at this point.
  • A value proposition links to the knowledge crystal: reducing covid-19 risk has some utility per a change in probability. There is no need to quantify this at the moment.

Now, the arguments in Kialo are less contextually positioned than when using this insight network, which consists of an action, two knowledge crystals, and a value proposition. For example,

  • "People's individual choices should be preserved" [4] is a value proposition attacking the action.
  • "Public safety, above all else, should determine whether wearing a mask is required" [5] is a value proposition about the knowledge crystal on the probability of disease.
  • "One size fits all is both ideologically and practically dangerous" [6] seems to be two separate arguments: the ideological part attacks the action, and the practical part attacks something else, probably the fraction of users in some way.

See also


  1. Wikipedia. Pareto efficiency. [1] accessed 2020-09-08
  2. 2.0 2.1 Wikipedia. Efficient envy-free division. [2] accessed 2020-09-08
  3. Kialo. Do people have a right not to wear a mask in public spaces during the covid-19 pandemic? [3] accessed 2020-11-05.