Skip to content
Discuss an evaluation

Indicant / Decision integrity for defense

Decision integrity for autonomous warfare

Indicant generates operational assessments and connects them to evidence and reasoning. Trace changes in their support to inform decisions as new information arrives.

Discuss an evaluation Explore the assessment workspace

Evidence, reasoning, and human judgment

Mineral terrain with distant aircraft and ground vehicles across a broad divide
Operational environmentExplore the assessment workspace ↓

01 / Generate an assessment

Generate assessments from the evidence

Indicant turns source footage and entity observations into situation assessments. Each conclusion stays connected to the evidence that supports it, with missing output, weak support, and unresolved identity visible.

How decision integrity works
A generated assessment with its evidenceIllustrative product view
Indicant evidence review at source 00:39.800, with the source frame, five candidate records, a generated scene assessment, and selected S10-V04 showing low-quality mask support. S10-SM01 is lost while three vehicle assessments retain support.
Source evidenceSelected S10-V04
Retained source frame at 00:39.800: vehicles and smoke on the roadway, with the original candidate mask overlays.
Source cutoff00:39.800
Selected S10-V04 support: low-quality output at source 00:39.800, its mask-quality reason, unassessed motion claim, and weakened assessment at 00:39.600.Generated scene assessment reports three visible vehicles and two tracks needing review. The queue distinguishes S10-V04 low-quality support from S10-SM01 missing output.

At source 00:39.800, Indicant generates a scene assessment from vehicle observations and their output states. Three vehicles retain screen-space motion support; missing and low-quality output remain explicit for two other candidates.

Open the complete workspace image
Read the evidence summary
Source and scene assessment
Source 00:39.800. The generated scene assessment reports three visible vehicles, two individual tracks needing review, and three mask centers moving right on screen.
Separate review reasons
S10-SM01 has no output at the selected source time; its assessment weakened at 00:39.700. S10-V04 is low quality because its mask is empty or its SAM score is below the declared quality floor; its assessment weakened at 00:39.600.
Retained support
S10-V03, S10-V01, and S10-V02 remain tracked. Their reported mask centers move right on screen, with assessments strengthened at 00:39.667.
Scope of the observation
Three candidate masks are shown at the selected source time. Mask quality, physical motion, and identity remain unverified. Candidate records, visible objects, and available masks have different scopes.

02 / Connect evidence to decisions

Keep decisions connected to their basis

Follow a conclusion through its claims to the source material. Compare competing explanations and see how later evidence changes support, while retaining what was known when the assessment was made.

Explore evidence and reasoning
The evidence behind a weakened assessmentIllustrative product view
Indicant entity review at source 00:11.900, with seven candidate records across debris, scooter, and vehicle groups. Selected S04-V02 is lost, with an assessment weakened at 00:11.733 and unresolved explanations for its missing output.
Source evidenceSelected S04-V02
Retained source frame at 00:11.900: vehicles, scattered debris and a scooter, with the original candidate mask overlays.
Source cutoff00:11.900
Selected S04-V02 evidence, claim and assessment support. No output is available at source 00:11.900; the assessment weakened at 00:11.733 and the cause remains unresolved.Source-time revision detail: S04-V01 and V03 revised at 00:11.567; S04-V02 weakened at 00:11.733; replay cutoff 00:11.900 shows five mask candidates. There are 93 scene revisions available.

At source 00:11.900, S04-V02 has no output and its assessment has weakened. Inspect the recorded reason alongside the source frame and revision history; the cause of the missing output remains unresolved.

Open the complete workspace image
Read the evidence summary

The evidence behind a weakened assessment

Selected candidate
S04-V02 is the white SUV partly cropped at the upper-left in the far lane. At source 00:11.900 it is lost, with its assessment weakened at 00:11.733.
Evidence, claim, and assessment
The selected SAM object has no output at this source time. Occlusion, exit, and model failure are not distinguished, so the available output does not establish physical movement or disappearance.
Context retained alongside the selection
The register contains three debris candidates, one scooter candidate, and three vehicle candidates; five candidate masks and 93 available scene assessment revisions describe the separate scopes of output and retained history.
Other supported observations
S04-V01 has mask-center movement left on screen and S04-V03 has movement up, with both assessments revised at 00:11.567. S04-D01 is lost while S04-D03 and S04-D02 remain tracked.
Follow support as the situation changesIllustrative product view
Indicant temporal review compares source 00:15.300 and 00:16.233, retaining five S05 candidate records and the shot boundary at 00:15.200. S05-V04 loses output while S05-V02 retains stationary screen-space support.
Source evidenceSelected S05-V04
Earlier retained source frame at 00:15.300: vehicles beneath the bridge and debris on the pavement, with candidate mask overlays.
Earlier cutoff00:15.300
Later retained source frame at 00:16.233 from the same S05 view, with candidate mask overlays.
Later cutoff00:16.233
S05-V04 support changes from tracked with insufficient motion support at 00:15.200 to lost and weakened at 00:15.933. S05-V02 instead gains stationary screen-space support at 00:16.000.The preceding shot boundary at 00:15.200 leaves identity across views unconfirmed. Both compared cutoffs belong to S05; their source and generation clocks remain distinct, with 126 earlier history revisions.

Compare retained S05 records at source 00:15.300 and 00:16.233. Indicant preserves the earlier basis while showing which assessments gained support, which weakened, and where identity remains unconfirmed.

Open the complete workspace image
Read the evidence summary

Follow support as the situation changes

Boundary and retained scene history
At source 00:15.300 the register contains five tracked S05 records and 126 scene assessment revisions are available. The 00:15.200 scene assessment marks a shot boundary and leaves identity across views unconfirmed.
Selected change in support
S05-V04 is tracked in the earlier register, with a created assessment at 00:15.200 and insufficient motion support. At source 00:16.233 it is lost, with its assessment weakened at 00:15.933 because the selected object has no output.
Support that remains
S05-V02 is tracked in both source views. Its earlier visible assessment is created at 00:15.200 with insufficient motion support; later it has no material screen-space displacement and its assessment is strengthened at 00:16.000.
Earlier context remains distinguishable
The retained 00:15.133 and 00:15.100 scene assessments report seven tracks needing review before the new view. The boundary does not establish that differently named candidates across those views share an identity.

Indicant informs operational judgment. Decisions and command authority remain with the responsible people and systems.

03 / Put Indicant to work

Evaluate Indicant on an operational question

Choose a question your team needs to answer. Compare the assessments Indicant generates, the evidence it connects, and the changes it surfaces against the way your team establishes the basis for a decision.

Start with one operational question, authorized evidence, and a named decision owner.

Discuss an evaluation Plan the evaluation