New thoughts and discussions

This page is intended for sharing your latest Intarese related thoughts and discussion topics that are already worth disseminating among interested people, but that might not be ripe enough to deserve their own pages yet. Feel free to write and process your thoughts here, no matter how unorganized they may be. Please edit the page so that the latest thoughts and discussions come on the top of the page.

11.9.2006 Connections between WP1.1-WP1.5

Participants: Jouni (motivated by previous discussions today with Matti and Juha)

Motivation: There should more interactions between SP1 WPs. What could this be?

• There should be methodological guidance for SP3.
• There should be collaboration related to scientific research within WP1.X, such as
  • intake fraction
  • Bayesian methods
  • expert elicitation
  • coherent analysis of uncertainty

Intake fraction generator

This tool has been invented by Matti, and here I try to operationalise it in Mediawiki as an example of Mediawiki's flexibility.

You want to start with the intake fraction definition:

iF = ∑ci*Pi*BR/Q,

where c=concentration, P=population, BR=breathing rate, Q=emission, i=index for location.
The idea is that you go through the equation step by step and choose the relevant values or models for calculating the values. This can be done in Mediawiki in the following way:

There should be a page named "Intake fraction generator". This page should contain the description of the process, and descriptions of all variables used in the equation. The possible values are described as a list of values or links to more detailed descriptions. When more text is needed to describe which value is appropriate, the links go to the relevant pages.

A new variable is created for the new intake fraction to be estimated. The focus and scope are described. The relevant variable values are copy-pasted from the description pages to the Input attribute. The Definition is of the form: iFgenerator(var1|var2|var3...) where iFgenerator is a function (a special kind of variable) that calculates the intake fraction based on the variables var1, var2, var3...

What wikipedia does is that it takes the code from iFgenerator Definition, replaces the Inputs with var1 and so on, and runs that code in R. The output is printed on a special page "Variable results", which contains results for variables where it has been possible to calculate the result. The result can be copy-pasted from there to the Result attribute of the new intake fraction variable.

A tool can be created that collects the information about all variables used, so that it can be printed out as documentation of the intake fraction estimate. Although we previously talked about Excel as the output format, this can now be forgotten, as the functionalities and documentation can be performed within wiki (using some extensions that are under development).

What this all means

If we are able to describe a process in risk assessment (such as intake fraction estimation) as variables with explicit definitions that only contain R code (including references to other variables), we can create a wiki page system that actually calculates results of such processes given the inputs. As all parts ot the process are wiki pages, they can be edited and improved freely within the system. Editing requires some experience, but when similar processes have been performed, there is an accumulating body of examples and previous estimates.

30.8.2006 Variable classes and discussion structure

Participants: Jouni

I started to think that maybe one attribute should be added to variables: Class. The possible classes could be

• Variable. The typical case with a physically measurable thing.
• Function. A procedure for calculating or evaluating the value of other variables. This would require parameters that are inputs for the function so that the value could be evaluated. Function class is necessary because functions will have an obligatory attribute parameters, and they cannot have variable result.
• Preference. The difference with a variable is that the result of a preference is measured by voting, while usually the result is measured by scientific observations or logical inference. On the other hand, also voting can be seen as a special case of scientific observation methods, and therefore
• Argument. It is not clear, whether this should be an own class, or whether argumentation will be held on talk pages.

Then I drafted some rules for argumentation:

• Statements can relate to any part of a variable.
• The argumentation must be located on the Talk page of the article page where the variable is located. (If the variable is moved to another page, also the argumentation must be moved.)
• The argumentation must be preceded by a) a subtitle with the same name as the variable and b) a sub-subtitle with a name describing the statement.
• At the part of the variable, about which the statement is, there must be a link to the talk page: [[Talk:{{PAGENAME}}#Sub-subtitle of the statement]].
• All arguments must be inside either {{Defend|argument}} or {{Attack|argument}} template.
• Arguments are structured hierarchically using indent with ':'.
• When an argumentation has come to a resolution, the outcome is described, and all other argumentation is hidden with {{Resolution|statement|outcome|argumentation}} template. The changes are made to the variable accordingly. Note! This template does not yet exist, because this version of Mediawiki does not support hierarchical templates.

25.8.2006 Refined ideas about pyrkilo object classes and their attributes

(also see below for previous version of the ideas and related links)

Participants: Jouni, Mikko

The only necessary object class is:

• Variable

The list of attributes that are needed to describe the variable object are:

• Name
• Focus
• Scope
• Description
• Inputs
• Index
• Definition
• Unit
• Result
• References

Function is a special case of a variable having some certain special characteristics (in practice parameters attribute), but is in its essence similar to other variables and thus belongs (and holds its place) in the variable class.

23.8.2006 Ideas about pyrkilo object classes and their attributes

Participants: Jouni, Mikko

The necessary object classes are:

• Variable
• Function

The needed attributes (equal to both object classes) are:

• Focus
• Scope
• Description (less formal definition)
• Definition (numerical/mathematical definition)
• Inputs (also Outputs, but this is not required since other attributes already define this)
• Unit
• Background information (prior)
• Result
• Index (list of independent entities)
• References (related external literature)

What is new/special about the lists above:

1. The attributes which values are always required are focus and scope, values for other attributes can be left unspecified if they are not relevant for the certain variable
2. Links are not considered as an individual object class anymore, instead links are defined as attribute values of variables
3. Function is a new individual object class which can be be called / pointed to with given parameters in the definition of a variable

Also see Glossary of environmental health#Variable attribute names and Risk assessment on ozone (AOT60) in Europe for more on these issues.

17.8.2006 An idea about combining Analytica and Wiki

Participants: Jouni

Can Analytica models be directly converted to wiki format?

• The code can be read from the xml file automatically.
• Variables in Ana can be transformed into Var templates
• Modules are created as categories
• When a node (variable) is located in a module, in wiki it means that the var template is categorized by the category defined by the module
• Nodezise, Nodelocation, and some Analytica-specific attributes are ignored in the transformation process.
• The wiki variable template must then contain the identical attributes: identifier, unit, title, description, definition, reference.
• Tables are difficult to transform automatically, but this problem needs not be solved in the first phase.

10.8.2006 A suggestion for the basic question in pyrkilo method

Participants: Jouni

The basic question: What are the rules that enable an open (non-organised, non-fixed) group of rational actors to describe environmental health risks and resolve disputes that arise during the process about the content?

Examples related to climate change

At least the following groups of rules are needed:

• Rules about participation (reading, editing, creating, and moving pages)
  • Clear: the 10 writers and the moderator may participate
• Rules about roles of participants (moderator, participants)
  • Clear
• Rules about structure (variables, links)
  • Structure drafted in Analytica; this is used as a strawman
• Rules about drafting and fixing the focus and scope
  • There are two different foci:
    1. description of climate change (this is about the world)
    2. description of meta level work about applying these rules and doing the assessment
  • Scope: All material from the 10 writers. Also additional material may be added by the moderator if necessary to make implicit issues explicit.
• Rules about acceptability of contribution (relevance, consistence etc., see below)
  • everything that is said in the material is relevant.
  • When material is conflicting (eg. Ahlbeck: CO2 does not increase global temperature), a higher-level statement is created: "There are conflicting opinions about CO2 and temperature"
• Rules about validation (argumentation, falsification)
• Rules about reorganising contributions (fusion, separation, budding)
  1. Make a strawman about the subject as a whole (after reading all material)
  2. Take the text from each writer and make it fit with the strawman. Reorganise the variables and links in the text as necessary. If there are variables in the text that do not yet exist in the strawman, add them there.
  3. Look at each piece of contribution and the existing strawman together. Make them fuse together either by editing the existing text in the strawman, or creating a new subvariable/link or supravariable/link as necessary, or splitting the variable into two. It is important to describe the relationship between the existing piece and the addition.
  4. After editing the variable, check for the consistence of
     1. incoming links
     2. outgoing links
     3. hierarchical relations up and down
  5. If needed, split or merge text into wiki pages as necessary
• Rules about resolving disputes
  • Disambiguation
  • Voting for values
  • (There is no need to resolve then in this example)

19.7.2006 Trying to define the objectives of pyrkilo method

Participants: Juha, Mikko, Jouni

The overall methodology for environmental health risk assessment should promote (or even restrict) the examination to have several high-level properties. These, and the tools and method that help to achieve the objective, are described below. The method that we develop in KTL is called the PYRKILO method, and we suggest that this, or selected parts of it, are taken into the INTARESE method as well.

Evaluating the desirability of outcomes
There is a need to perform risk assessment only if some of the possible outcomes are more desirable than others. What is desirable and what is not, is a value judgement. These can be resolved using democratic methods such as voting.

Describing the situation in a rational way
The assessment must be rational and inherently consistent. Logic and directed acyclic graphs (DAGs) are the tools to promote this.

Evaluating relevance of issues
The individual pieces must be relevant for the whole assessment. This can be evaluated using argumentation theory.

Evaluating whether the issues in the assessment are probable
Low probability issues are less important for the assessment, and the tool to evaluate this is probability theory, especially Bayesian statistics, which offer some very nice tools when combined with DAGs. It should be noted, however, that 'low probability' is not a fixed number: the collision of a meteorite to the Earth may have a very low probability, but it may still be important due to its potential impact (see utility).

Evaluating the utility of actions
The utility of the outcome is a key indicator for decision-making.

Evaluating the importance of uncertainties
A value-of-information analysis is needed to estimate, whether a particular uncertainty makes it difficult to decide in a particular context.

Overall, the new pyrkilo method should have such a structure and process rules that it facilitates assessments that are in line with all of these six objectives.

• Desirability - democracy
• Rationality - logic, DAGs
• Relevance - argumentation theory
• Probability - probability theory
• Utility - decision theory
• Importance of uncertainties - value-of-information analysis

How does the first version of the pyrkilo tool look like?

• It is a systematic collection of relevant and structured information.
• It does not compute anything. All computing is done with other tools, and results are uploaded to the tool.
• It is based on Mediawiki program.
• It is an open-access system, but for a limited group of people. In this sense, it could look much like this Intarese Wiki site.
• It has a prespecified structure for assessments, variables, and variable attributes.
• It has a prespecified categorisation systems for a) variables, b) tasks and processes.
• It has procedural rules for how to link variables to each other.

--Jouni 14:23, 19 July 2006 (EEST)

4.7.2006 thoughts after the ISSA argumentation conference

Computer aided or web-based argumentatio applications to look into:

• Belvedere (?)
• Questmap (http://www.compendiuminstitute.org/tools/questmap.htm)
• Argumentum (http://www.argumentum.ch)
• Argunet (http://www.denkartist.de/argunet)
• more?

Some names with potentially useful ideas/research on argumentation (pragma-dialectics, epistemology...):

• Gábor Kutrovátz, Eotvos University of of Budapest
• Ralph H. Johnson, Univeristy of Windsor
• Dan Cohen, Colby College
• Robert C. Rowland, University of Kansas
• Michel Dufour, University Sorbonne Nouvelle
• Christoph Lumer, University of Siena
• Fabio Paglieri, University of Siena

Co-operation potential???

• Gábor Kutrovátz, Eotvos University of of Budapest
• Sara Rubinelli, University of Lugano
• David M. Berube, University of South Carolina NanoCenter

Mikko 16:59, 4 July 2006 (EEST)

21.6.2006 technical issues in using Wiki

Mikko Pohjola

To make the Wiki working environment clear, useful and effective we must agree on certain kind of code of conduct and build the system in a way that it guides the users to right kinds of action in using it. In brief it is a question of:

• effective design of the system
• good user guidance to
  • the technical questions
  • use of the method
• (minimal) control of the system

The first bullet, design, actually includes the second bullet, because the guidance is given in/by the system. Also the ways to minimize the need of control are in the design of the system. Since the environment is fixed in using Wiki, the design then means (a) how to make best use of the existing properties of Wiki (b) for the needs of environmental health risk assessment. Therefore the design is first methodological and second technical.

The methodology is now starting to take shape (Help:Writing pyrkilo risk assessments) in a way that it is about the time to take the technical issues into more detailed scrutiny. Open questions concerning this that require consideration in near future are at least e.g.:

• use of categories to group the pages
  • operational environment -related
  • substantially relevant
  • managerial (project management etc.)???
• use of templates to "standardize" the appearance
  • attribute tables
  • pre-framed pages
  • sharing content
• control of hierarchy???
• user guidance
  • technical
  • methodological

1.6.2006 On argumentation analysis for risk assessment

Participants: Jouni & Mikko

• the role of argumentation analysis in risk assessment is to fill the gap between explicit modelling/calculation and unstructured discussion
  • helps e.g. in making complicated models easier to intrepret by non-experts and brings out the essentials of a discourse
• the primary goal concerning argumentation analysis' utilization for risk assessment is to make it a common, easy-to-use, a priori method for risk assessors
  
• also use of argumentation theory as an a posteriori method for reconstructing discourses afterwards can be fruitful
  • this line of use is probably necessary in explicating the usefulness of the method
    
• discussion in Science about health risks/benefits of salmon shall be made an example for using argumentation analysis in risk assessment
• in this example the assessor (as an "outsider") chooses the focus and scope of the argumentation analysis
  • if the basic standpoint of the analysis, and thus the point of view, is chosen solely based on the first article by Hites et al., it would be difficult to fit all the replies in the same picture (the disagreement space varies in settings Hites vs. Rembold, Hites vs. Tuomisto, etc...)