Competency Analyzer

Observational research software

Studying how people learn to work on sustainability problems together.

Competency Analyzer is a desktop research instrument for the systematic observational analysis of recorded collaborative learning. It segments a session into interactional units, applies a researcher-defined codebook to each one, and reports counts that remain traceable to the verbatim talk they came from.

Download for Windows Coming soon Read the method

Pre-release. The application is under active development and is not yet distributed. This page will host the installer when a release build is available.

Approach

Evidence about conduct, not claims about people.

Sustainability problems are worked out in conversation — across disciplines, languages and value positions. This instrument is built to describe that conversation as it can be observed, and to stop there.

Observable conduct

Coding is applied to what the transcript and video show: what was said, how turns were taken, what was repaired, what was displayed. Internal states, intentions and abilities are not inferred.

Verbatim traceability

Each coded entry retains the words it was drawn from, so a reader can compare the talk behind a count in one session against another rather than reading the count alone.

Configurable codebook

The competency areas, the affordance vocabulary and the argumentation categories are all configuration. Researchers supply the definitions their study will defend.

Plural contexts

Argumentation is coded for own-context framing, bridging between positions and value divergence — treating difference as material to be described, not deviation from one norm.

Researcher authority

Any automated code can be overridden, with reviewer initials, a note and a timestamp retained alongside the original value.

Data stays local

Audio can be transcribed on the researcher's own machine, so recordings need not be transmitted to any external service to produce a transcript.

Coded areas

Competency areas in the default configuration.

These are descriptive category labels shipped as an editable default. Their definitions are set by the researcher operating the software, to suit the study they are used in.

Categories are counted, not scored. A count records how often talk was coded into a category in a given session; it is not a measurement of an individual's ability, attainment or development.

Method

How an analysis proceeds.

  1. Preparation

    A recording is added to a workspace. Audio extraction and frame sampling run locally on the researcher's computer.

  2. Transcription

    Speech is transcribed either on the local machine or through a configured hosted deployment. Speaker attribution is recorded together with the method that produced it, so the share of talk actually identified can be reported rather than assumed.

  3. Segmentation and unit coding

    The session is divided into interactional units of roughly thirty to sixty seconds. Each unit is coded independently, with a bounded window of preceding talk available as read-only context.

  4. Session-level reading

    Some patterns appear only across a stretch of talk, never in a single turn. A second pass reads the transcript in windows that overlap at the edges, so a sequence sitting on a boundary is still seen whole by one of them. Repeats from those shared edges are collapsed, and a finding survives only if the words it quotes appear in the transcript.

  5. Review

    Coded units are presented on a timeline with their evidence, for inspection and override before anything is exported.

  6. Reporting

    Session reports, a codebook glossary and change-over-time reports across an ordered set of sessions are exported to Excel and PDF, with participation distribution and repair frequency reported alongside the counts.

Scope & limits

Scope of the instrument.

Stated plainly, because a research instrument is only as useful as its declared boundaries.

It does not measure ability

Outputs are frequency counts of coded conduct. They are not test scores, competence ratings, grades or evidence of individual attainment, and must not be used for assessment of individuals.

It does not read emotion

Only observable displays are coded, from visible configuration, posture, laughter and affect-laden wording. Vocal pitch and volume are not analysed. No inference is made about what anyone felt.

Warrant rests with the study

The software supplies coded observations and the evidence behind them. Their standing as findings is established by the research design they serve and by the researcher who reviews them.

It does not identify people

Reports use pseudonymous speaker labels. Attribution can be incomplete, and the proportion of talk left unattributed is reported so a quiet participant is not confused with a missed one.

Ethics & data protection

Responsibilities of the operating researcher.

Recordings of people are personal data. Anyone operating this software is solely responsible for establishing a lawful basis for processing, for obtaining informed consent from participants, for securing any required ethical review or institutional approval, and for complying with all applicable data protection and research governance requirements in their jurisdiction.

Nothing on this page constitutes legal advice, nor a representation that use of the software is lawful in any particular setting. The software is provided for research use, without warranty of any kind, and no fitness for any particular purpose is asserted.

No affiliation with, endorsement by, or approval from any institution, funder, publisher or third-party organisation is claimed or implied. All third-party names, standards and software referred to in the application remain the property of their respective owners.

Availability

Not yet released.

The application is in development and is not currently distributed. When a release build is published, the installer will be linked from this page.

Download for Windows Coming soon

Windows 10 or later. Access is granted by written agreement with the rights holder.