SACM 1.1 RTF Avatar
  1. OMG Issue

SACM11 — Uncertainty characteristic for claims

  • Key: SACM11-3
  • Legacy Issue Number: 16287
  • Status: closed  
  • Source: MITRE ( Mr. Samuel Redwine)
  • Summary:

    From the viewpoint of compatibility with ISO/IEC 15026, the most important issue is that SACM contain an "Uncertainty" characteristic or attribute, a key aspect of 15026-2, in order to facilitate use of SACM with ISO/IEC 15026 conformant assurance cases. These would be useful throughout SACM as attributes of assertions or claims (not relations).

    Current usage means a need exists for several data type options ­ not necessarily mutually exclusive.

    Data Types
    • Probability true: 0-1.00
    • Type I error: 0-1.00
    • Type 2 error: 0-1.00
    • Enumeration: None, low, medium, large, unknown
    • Standard deviation: real number
    One might also want to include a data type with customizable meaning (values: 1-100) that would parallel the data type for “Strength” in present draft if this is retained.

    This attribute element is associated with claims (assertions) and not with relations. Current SAEM draft defines characteristics ­ such as “Strength,” and “Confidence” ­ of evidence relations between an EvidenceItem (e.g. exhibit) and a DominAssertion. These relate to uncertainty but are associated with a relation and not with a DomainAssertion (claim). Note that the draft of SAEM states these are one-to-one relations unlike the many-to-one inferential relations in ARM.

  • Reported: SAEM 1.0b1 — Fri, 27 May 2011 04:00 GMT
  • Disposition: Deferred — SACM 1.1
  • Disposition Summary:

    Deferred as still valid for SACM

    Whilst still a gap between what is discussed in ISO 15026 and what is supported in SACM, it was recognized that there still is no common established state of the practice for recording confidence and uncertainty values on assurance case argument elements. It may be that a probability value would allow for the most commonality amongst current approaches. Guidance would be required on how to map to a probability value under different confidence determination regimes as listed above. There was still some disagreement about whether the current approach to modeling confidence in the evidence portion of SACM is fit for purpose. However for backward compatibility now changes have been made to this.

  • Updated: Tue, 21 Apr 2015 01:16 GMT