Skip to content

Add Hypothesis (statement) and EvaluationResult (AnalyticResult) #190

Description

@eoghanscasey

Background

Investigations involve hypothesis generation and evaluation of observations in light of competing hypotheses. Evaluation assigns a value to observations in light of each hypothesis that can be fed into a decision process. As described in NIST OSAC Tech Pub 0002, the value produced by evaluation can be a measure of continuous nature, such as a real number between 0 and 1 (probability), or a real number between -1 and 1 (measure of distrust and trust). Alternatively, this value can a mark of a discrete nature, such as {A+, A, A-, B+, B, B-, C,D, F}... or a verbal scale.

Requirements

Requirement 1

State a Hypothesis and one or more alternative Hypotheses being considered. Selecting a single hypothesis without consideration of alternatives increases risk of confirmation bias.

Requirement 2

Evaluate observations under multiple hypotheses and assign a value (numeric, verbal, etc.) to the observations, not the hypothesis.

Requirement 3

Allow for different forms of evaluation, including verbal (more probable), numeric (0.9), pseudonumeric (C5.5)

Requirement 4

Represent who performed the evaluation, using which tools and methods.

Requirement 5

Group a set of hypotheses being considered

Risk / Benefit analysis

Benefits

The ability to generate and evaluate hypotheses is fundamental to cyber-investigations.

Risks

The analysis of competing hypotheses approach represented here are incompatible with conclusion scales (e.g., Identification, Inconclusive, and Exclusion), which typically only consider one hypothesis (problematic because of confirmation bias), are categorical (problematic because of lack of agreement and cliff-edge effect), and do not make a clear distinction between hypotheses evaluation (prosector's fallacy) and evaluation of observations in light of opposing hypotheses.

The evaluative approach represented here is incompatible with non-scientific domains (e.g., intelligence analysis) approaches that assign a likelihood of occurrence of an event or development, and the confidence in the basis for this judgment. Assigning a value to an analytic judgment evaluates the hypotheses not the observations (prosector's fallacy), increases risk of confirmation bias, and is not balanced with evaluation of opposing hypothesis.

Competencies demonstrated

Competency 1

Evaluate the observations given opposing hypotheses.

Competency Question 1.1

Are the [Observations] more probable given [Hypothesis 1] or [Hypothesis 2]?

Result 1.1

The [Observations] are more probable given [Hypothesis 1] rather than [Hypothesis 2].

Competency Question 1.2

What [Observations] were evaluated under [Hypothesis 1] and [Hypothesis 2]?

Result 1.2

The [Observations] were evaluated under [Hypothesis 1] and [Hypothesis 2].

Competency 2

Track provenance and ensure accountability for evaluation of observations given opposing hypotheses.

Competency Question 2.1

Who performed the evaluation of the observations given opposing hypotheses?

Result 2.1

Eoghan Casey performed the evaluation of the observations given opposing hypotheses.

Solution suggestion

Add Hypothesis (Assertion): A statement that is either true or false, the truth of which is uncertain

Add EvaluationResult (AnalyticResult): An opinion that is formed on the basis of observations in light of a given hypothesis.

Add HypothesisGroup (Collection): A collection of hypotheses being considered.

This illustrative example demonstrates Requirement 1:

https://github.com/casework/CASE-Examples/tree/ONT-434/examples/illustrations/inference

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions