Data Affect LLC · Research & PracticeComputer & information science research · Ashburn, Virginia
Data Affect
Data Affect / Research program

Research program

How we conduct and document research.

Our research asks questions about software and information that we could not answer at the outset. This page sets out what the program covers, the method we use, how each program is written up against the four-part test for research activities, and how research is kept apart from client work.

Programs
P-01 Symbol Grounding Framework
P-02 Recova (research preview)
Field
Computer science; information science
Framework
Four-part test, IRC §41(d)1

1Scope

The program is concerned with problems in computer science and information science that arise when organizations try to make their information dependable: for people, for software, and for AI systems. It currently has two programs.

  • P-01 · Symbol Grounding Framework (since 2024). How business terms, documents and rules can be represented so that information carries clear meaning and a traceable source. Program description.
  • P-02 · Recova (research preview). How records retention, legal holds and privacy obligations that apply to the same information can be represented, versioned and evaluated together in one software architecture. Program description.

2The four-part framework

Each program description is organized under the four parts of the test for research activities in Internal Revenue Code §41(d).1 We use the framework because it asks the questions a careful technical write-up should answer anyway. Using it is a documentation choice, not a claim about how any activity is treated for tax purposes.

PartThe question it asksWhere it appears
1Permitted purpose. Is the work aimed at a new or improved function, performance, reliability or quality of a business component, such as software or a process?“Objective” section of each program
2Technological in nature. Does the work fundamentally rely on principles of the physical or biological sciences, engineering, or computer science?“Technical basis”
3Elimination of uncertainty. At the outset, was there uncertainty about the capability or method of achieving the result, or about the appropriate design?“Research questions”
4Process of experimentation. Were alternatives identified and evaluated, through modeling, simulation, or systematic trial and error, to resolve the uncertainty?“Experimentation” and the research log

3Method

Work proceeds in iterations. Each begins with a stated question and a hypothesis about how the software should behave, compares more than one way of achieving it, builds the smallest change that can test the idea, and runs it against controlled scenarios, comparing the result with the behavior expected beforehand. The observation decides what happens next: the hypothesis is accepted, the design is refined, or the approach is abandoned.

PROCESS OF EXPERIMENTATION SCHEMATIC 01Questionthe uncertainty 02Hypothesisexpected behavior 03Alternativesdesigns compared 04Prototypesoftware change 05Test & observecontrolled scenario 06Resultaccept · refine · reject ITERATE: REFINE THE HYPOTHESIS OR THE DESIGN EACH STEP DATED IN THE RESEARCH LOG
FIG. 2The experimentation cycle used to describe each program. A result either closes a question or sends the work back for another iteration. Schematic.

Where results are classified, we separate defects, where the software did not do what the design intended, from design limitations, where the design itself turned out to be wrong or incomplete. Design limitations are architectural findings, and the resulting design changes are what the program descriptions report.

4What each record contains

Each iteration of the work is described using the same elements, and the research log summarizes milestones by date:

  1. a.Date and programWhen the work happened, and which program it belongs to.
  2. b.QuestionThe specific uncertainty being addressed.
  3. c.Hypothesis and alternativesWhat was expected, and which designs were compared.
  4. d.TestThe prototype or change, and the scenario it was run against.
  5. e.Observation and decisionWhat happened, and whether the design was kept, refined or dropped.

The public log is a summary; it does not publish detailed technical records.

5Research and client work

Data Affect also does consulting. The two are kept apart, because they are different kinds of work.

Research program

  • Questions chosen and directed by Data Affect
  • Outcome not known in advance; work may fail
  • Described on these research pages and in the log

Client consulting

  • Scope agreed with, and paid for by, the client
  • Applies established methods to the client’s needs
  • Summarized under Practice
  • Not presented on this site as research

Research paid for by another party, such as under a client contract or grant, is generally excluded from the research credit.2 Whether any particular activity or cost is treated as qualified research is decided separately, with our tax advisers, and is not stated on this site.

6Notes

  1. Internal Revenue Code §41(d) and Treasury Regulation §1.41-4. The four parts are commonly summarized as permitted purpose (the business-component test), technological in nature, elimination of uncertainty, and process of experimentation. They are used here only to organize the descriptions. ↩
  2. IRC §41(d)(4)(H), “funded research”. ↩
  3. This page is not tax, legal or accounting advice. See the note at the foot of every page.

Correspondence

Questions about the method?

We are glad to discuss how the work is organized.