Research & Constraints

Problem Class Under Study

Automated and AI-driven systems frequently escalate under uncertainty. When signals are incomplete or ambiguous, many systems default to action rather than restraint.

This creates a failure mode: once execution begins, escalation is often irreversible. Premature actions compound, and downstream systems inherit flawed assumptions.

TrigGuard exists to study and enforce non-action as a safety response. It treats refusal as a valid system output, not a fallback or error state.

What This Research Does Not Claim

The following limits are intentional and define the boundaries of this work:

• No claims of behavioural improvement in users
• No claims of outcome optimisation or performance gains
• No claims of intelligence, prediction, or reasoning capability
• No claims of autonomy or agency
• No therapeutic, medical, or diagnostic claims

Validation and Review Methods

TrigGuard validation focuses on process integrity rather than outcome metrics:

• Deterministic simulation testing Execution logic is tested against controlled input sets with known expected outputs.
• Shadow-mode deployment The system observes real execution contexts without taking action, allowing validation without consequence.
• Adversarial and edge-case input testing Inputs designed to exploit ambiguity or trigger unsafe behaviour are systematically tested.
• Internal ethics and safety review Design decisions are reviewed against harm-reduction principles before implementation.
• External review Where appropriate, external reviewers assess system behaviour and constraint adherence.

Known Failure Modes and Limits

TrigGuard has known limitations that users and integrators should understand:

• False positives The system may over-block execution, refusing actions that would have been safe. This is a deliberate trade-off favouring safety over availability.
• Loss of contextual fidelity Under extreme ambiguity, the system cannot distinguish between edge cases. It defaults to refusal.
• Trade-offs between safety, availability, and responsiveness A fail-closed system is inherently less responsive. Speed is never prioritised over safety.

Research Status

TrigGuard currently operates in controlled, private beta and shadow-mode contexts only.

Research is ongoing. Public performance or outcome claims are intentionally constrained until validation processes are complete and externally reviewed.

Educational Resources

The following resources inform TrigGuard's design philosophy. They are educational, not diagnostic or prescriptive.

Neurodivergence & Emotional Regulation

Neurodivergent individuals, including those with ADHD, autism, and related conditions, often experience differences in emotional regulation and sensory processing. These differences are not deficits but represent diverse ways the brain processes information and responds to the environment.

ADHD is frequently associated with emotional dysregulation, including difficulty managing frustration, impatience, and mood variability. Autistic individuals may experience sensory overload when environmental stimuli exceed their processing capacity, leading to overwhelm and shutdown responses.

Understanding these experiences is essential for developing digital wellbeing tools and neuroadaptive systems that genuinely support rather than burden users.

Research References

The following resources from reputable institutions provide further reading on these topics: