Open assessment

This page is a encyclopedia article. The page identifier is Op_en2875
Moderator:Jouni (see all)

Upload data {{#opasnet_base_link:Op_en2875}}

For a description about a web-workspace related to open assessment, see Opasnet.

Open assessment is a method that attempts to answer the following research question and to apply the answer in practical assessments: how can scientific information and value judgements be organised for improving societal decision-making in a situation where open participation is allowed?

In practice, the assessment processes are performed using Internet tools (notably Opasnet) along with more traditional tools. Stakeholders and other interested people are able to participate, comment, and edit the content as it develops, from an early phase of the process. Open assessments explicitly include value judgements, thereby extending its application beyond the traditional realm of risk assessment into the risk management arena. It is based, however, on a clear information structure and scientific methodolgy in order to provide clear rules for dealing with disputes. value judgements thus go through the same open criticism as scientific claims; the main difference is that scientific claims are based on observations, while value judgement are based on opinions of individuals.

Like other terms in the field of assessment 'open assessment' is subject to some confusion. It is therefore useful to distinguish clearly between:

the open assesment methodology;
the open assessment process - i.e. the actual mechanism of carrying out an open assessment, and
the open assessment product or report - i.e. the end product of the process.

To ensure clarity, open assessment also attempts to apply terms in a very strict way. In the summary below, therefore, links are given to further information on, and definitions of, many of the terms and concepts used. <section end=glossary />

Open assessment as a methodology

Open assessment is built on several different methods and principles that together make a coherent system for collecting, organising, synthesising, and using information. These methods and principles are briefly summarised here. A more detailed rationale about why exactly these methods are used and needed can be found in the Open assessment method. In addition, each methodor principle has a page of its own in Opasnet.

Purpose

The basic idea of open assessment is to collect information that is needed in a decision-making process. The information is organised as an assessment that predicts the impacts of different decision options on some outcomes of interest. Information is organised to the level of detail that is necessary to achieve the objective of informing decision-makers. An assessment is typically a quantitative model about relevant issues causally affected by the decision and affecting the outcomes. Decisions, outcomes, and other issues are modelled as separate parts of an assessment, called variables. In practice, assessment and variables are web pages in Opasnet, a web-workspace dedicated for making these assessments. Such a web page contains all information (text, numerical values, and software code) needed to describe and actually run that part of an assessment model.

Basic concepts

These web pages are also called information objects, because they are the standard way of handling information as chunk-sized pieces in open assessments. Each object (or page) contains information about a particular issue. Each page also has the same, universal structure: a research question (what is the issue?), rationale (what do we know about the issue?), and result (what is our current best answer to the research question?). The descriptions of these issues are built on a web page, and anyone can participate in reading or writing just as in Wikipedia. Notably, the outcome is owned by everyone and therefore the original authors or assessors do not possess any copyrights or rights to prevent further editing.

The structure of information objects can be likened to a like fractal: an object with a research question may contain sub-questions that could be treated as separate objects themselves, and a discussion about a topic could be divided into several smaller discussions about sub-topics. For example, there may be a variable called Population of Europe with the result indexed by country. Instead, this information could have been divided into several smaller population variables, one for each country - for example in the form of a variable called Population of Finland. How information is divided or aggregated into variables is a matter of taste and practicability and there are no objective rules. Instead, the rules only state that if there are two overlapping variables, the information in them must be coherent. In theory, there is no limit to how detailed the scope of an information object can be.

Trialogue is the term used to define Wikipedia-like contributions. The trialogue concept emphasises that, in addition to having a dialogue or discussion, a major part of the communication and learning between the individuals in a group happens via information objects, in this case Opasnet pages. In other words, people not only talk or read about a topic but actually contribute to an information object that represents the shared understanding of the group. Wikipedia is a famous example of trialogical approach although the wikipedists do not use this word.

Groups are crucial in open assessment because all research questions are (implicitly) transformed into questions with the format: "What can we as a group know about issue X?" The group considering a particular issue may be explicitly described, but it may also be implicit. In the latter case, it typically means anyone who wants to participate, or alternatively, the whole of humanity.

The use purpose of information is crucial because it is the fuel of assessments. Nothing is done just for fun (although that is a valid motivation as well) but because the information is needed for some practical, explicit use. Of course, other assessments are also done to inform decisions, but open assessments are continuously being evaluated against the use purpose; this is done to guide the assessment work, and the assessment is finished as soon as the use purpose is fulfilled.

Open assessment attempts to be a coherent methodology. Everything in the open assessment methodology, as well as in all open assessment process, is accepted or rejected based on observations and reasoning. However, there are several underlying principles that cannot be verified using observations, called axioms of open assessment. The six axioms, which are essentially Cartesian in origin, are:

The reality exists;
The reality is a continuum without, for example, sudden appearances or disappearances of things without reason;
I can reason;
I can observe and use my observations and reasoning to learn about the reality;
Individuals (like me) can communicate and share information about the reality;
Not everyone is a systematic liar.

Basic procedures

Inference rules are used to decide what to believe. The rules are summarised here. 1) Anyone can promote a statement about anything (promote = claim that the statement is true). 2) A promoted statement is considered valid unless it is invalidated (i.e., convincingly shown not to be true). 3) The validity of a statement is always conditional to a particular group (which is or is not convinced). 4) A statement always has a field in which it can be applied. By default, a scientific statement applies in the whole universe and a moral statement applies within a group that considers it valid. 5) Two moral statements by a single group may be conflicting only if the fields of application do not overlap. 6) There may be uncertainty about whether a statement is true (or whether it should be true, in case of moral statements). This can be quantitatively measured with subjective probabilities. 7) There can be other rules than these inference rules for deciding what a group should believe. Rules are also statements and they are validated or invalidated just like any statements. 8) If two people within a group promote conflicting statements, the a priori belief is that each statement is equally likely to be true. 9) A priori beliefs are updated into a posteriori beliefs based on observations (in case of scientific statements) or opinions (in case of moral statements) and open criticism that is based on shared rules. In practice, this means the use of scientific method. Opinions of each person are given equal weight.

Tiers of open assessment process describe typical phases of work when an open assessment is performed. The tiers are the following: Tier I: Definition of the use purpose and scope of an assessment. Tier II: Definition of the decision criteria. Tier III: Information production. It is noteworthy that the three tiers closely resemble the first three phases of IEHIA, but the fourth phase (appraisal) is not a separate tier in open assessment. Instead, appraisal and information use happens at all tiers as a continuous and iterative process. In addition, the tiers have some similarities also to BRAFO approach, although the tiers in these approaches are not the same.

It is clear that within a self-organised group, not all people agree on all scientific or moral statements. The good news is that it is neither expected nor hoped for. There are strong but simple rules to resolve disputes, namely rules of structured discussions. In straightforward cases, discussions can be informal, but with more complicated or heated situations, the discussion rules are followed. 1) Each discussion has one or more statements as a starting point. The validity of the statements is the topic of the discussion. 2) A statement is valid unless it is attacked with a valid argument. 3) Statements can be defended or attacked with arguments, which are themselves treated as statements of smaller discussions. Thus, a hierarchical structure of defending and attacking arguments is created. 4) When the discussion is resolved, the content of all valid statements is incorporated into the information object. All resolutions are temporary, and anyone can reopen a discussion. Actually, a resolution means nothing more than a situation where the currently valid statements are included in the actual content of the respective information object.

Technical functionalities supporting open assessment

Opasnet is the web-workspace for making open assessments. The user interface is a wiki and it is quite similar to Wikipedia, although it also has enhanced functionalities for making assessments. One of the key ideas is that all work needed in an assessment can be performed using this single interface. Be it information collection, numerical modelling, discussions, statistical analyses on original data, publishing original research results, peer review, organising and distributing tasks within a group, or dissemination of results to decision-makers, it's all there and it's all available for anyone to use and participate. Opasnet is an overall name for many other functionalities than the wiki, but because the wiki is the interface for users, Opasnet is often used as a synonym for the Opasnet wiki. Other major functionalities exist as well, and they are presented next. The main article about this topic is Opasnet structure.

Most variables have numerical values as their results. Often these are uncertain and they are expressed as probability distributions. A web page is an impractical place to store and handle this kind of information. For this purpose, a database called Opasnet Base is used. It is a very flexible storage, and almost any results that can be expressed as two-dimensional tables can be stored in Opasnet Base. Results of a variable can be retrieved from the respective Opasnet page. Opasnet can be used to upload new results into the database. And finally, if one variable B is causally dependent on variable A, the result of A can be automatically retrieved from Opasnet Base and used in the formula for calculating B.

Because Opasnet Base contains samples of distributions of variables, it is actually one huge Bayesian belief network, which can be used for assessment-level analyses and conditioning and optimising different decision options. In addition to finding optimal decision options, Opasnet Base can be used to assess the value of further information for a particular decision. This statistical method is called Value of information.

Opasnet contains modelling functionalities for numerical models. It is an object-oriented functionality based on R statistical software and the results in Opasnet Base. Each information object (typically a variable) contains a formula which has detailed instructions about how its result should be computed, often based on results of upstream variables in a model.

Meta level functionalities

In addition to work and discussions about the actual topics related to real-world decision-making, there is also a meta level in Opasnet. Meta level means that there are discussions and work about the contents of Opasnet. The most common thing to see about meta level are the rating bars in the top right corner of many Opasnet pages. Peer rating means that users are requested to evaluate the scientific quality and usefulness of that page on an axis from 0 to 100. This information can then be used by the assessors to evaluate which parts of an assessment require more work, or by readers who want to know whether the presented estimates are reliable for their own purpose.

The users are also allowed to make peer reviews of pages. These are similar to peer reviews in scientific journals with written evaluations of the scientific quality of content. Another form of written evaluation is acknowledgements, which is a description about who has contributed what to the page, and what fraction of the merit should be given to which contributor.

Estimates of scientific quality, peer reviews and acknowledgements can be used to systematically calculate how much each contributor has done in Opasnet. These practices are not yet, however, well developed. Contribution scores are so far the only systematic method to even roughly estimate contributions quantitatively.

Respect theory is a method for estimating the value of freely usable information objects to a group. This method is under development, and hopefully it will provide practical guidance for distributing merit among contributors in Opasnet.

Why does open assessment work?

Most people think that it is unbelievable that open assessment works (i.e. they don't believe that it could work). So far, there are only small-scale evidence about its performance, showing things like "There exists people who have been able to find a suitable topic to make an open assessment that has converged to a conclusion without falling apart due to attacks (as there were no attacks to mention)." This is still very far (substance-wise) from conclusive evidence about my statement below, although I hope that time-wise we are not very far from the future where that conclusive evidence exists. I will now present my statement and then attempt to defend it with current understanding of making assessments, nature of information, and behavour of human beings.

Open assessment or approaches adopting similar principles will take over a major part of information production motivated by societal needs and improvement of societal decision-making. The main overall defence for this statement is economic: open assessment is cheaper to perform, easier to utilise, and of higher quality than current alternative methods to produce societally important information. There are several reasons to believe in this, at least that it would work with the most important issues.

In all assessments, there is a lack of resources, and this limits the quality of the outcome. With important (and controversial) topics, opening up an assessment to anyone will bring new resources to the assessment in the form of interested volunteers.
The rules of open assessment make it easy enough to organise the increased amount of new data (which may at some points be of low quality) into high-quality syntheses within the limits of new resources.
Participants are relaxed with the idea of giving away important information -- a prerequisite of an effective open assessment -- because open assessments are motivated by hope for societal improvements and not by monetary profit. This is unlike in many other areas where information monopolies and copyrights are promoted as means to gain competitive advantage in a market, but as a side effect leading to information clogs.
Especially problems due to too narrow initial scoping will become less common with more eyes looking at the topic already when it is assessed.
It becomes easier to systematically apply the basic principles of the scientific method, namely rationale, observations, and especially open criticism.
Any information organised for any previous assessment is readily available for a new assessment related to the same topic. The work time for data collection and also calendar time from data collection to utilisation get shorter. This increases efficiency.
All information is organised in a standard format which makes it possible to develop powerful standardised methods for data mining and manipulation and consistency checks.
It is technically easy to prevent malevolent attacks against the content of an assessment (on a topic page in Opasnet wiki) without restricting the discussion about and improvement of the content (on a discussion page related the particular topic page) in any way; the resolutions from the discussions are simply updated to the actual content on the topic page by a trusted moderator.

Keywords

Open assessment, Opasnet, variable, assessment, information object, statement, validity, discussion, group

References

Related files