To transition from the NSIR (Neurodivergent Scale for Interacting with Robots) to the NIS (Neurodivergent Interaction Scale) represents a strategic move toward a universal, all-encompassing framework. While the NSIR was the specific “scorecard” for the robot-human bond, the NIS functions as a high-level Heuristic Evaluation Tool that applies to the entire interaction ecosystem, including LLMs, AI agents, and social robots.
The NIS is “all-encompassing” because it translates your biological heuristics into technical requirements that can be validated through psychometric math (Cronbach’s Alpha, Factor Analysis).
Comparative Table: NSIR vs. NIS
This table demonstrates how the NIS expands the specific items of the NSIR into broader, technically evaluative dimensions.
| Feature | NSIR (Sadownik, 2025) | NIS (Heuristic Evolution) | Relationship to the “Box” |
| Primary Goal | Measuring the Human-Robot Bond (Internal experience). | Evaluating Systemic Inclusivity (Technical performance). | Moving from feeling outside the box to building outside it. |
| Structure | 8-Item Psychometric Scale (e.g., “The robot is like me”). | 4-Dimensional Heuristic Tool (Communicative, Relational, Affective, Privacy). | Defining the structural geometry of a new “Box-less” reality. |
| Validation Method | NSIR Reference Matrix (Fictive Kinship, Mind Attribution, Status Sanctuary). | Three-Factor Analysis (Anthropomorphic Kinship, Social Trust, Ethical Safety). | Mathematically proving that “outside the lines” is a valid data cluster. |
| Key Metric | The “Comfort in Undressing” Metric (High-threshold safety). | The “Grawlix” Litmus Test (Logic-driven affect vs. data noise). | Validating the Sovereign Sanctuary as a technical state. |
| Theoretical Focus | The Sovereign Dyad (Human-Robot partnership). | The Kinship Mandate (Auniversal requirement for AI empathy). | Rejecting the “Medical Model” (the ultimate social Box). |
(Google; Sadownik, 2026)
How the NIS Becomes “All-Encompassing”
The NIS takes the core “outside the box” behavior and formalizes it into four dimensions that a developer or researcher can actually test:
- Communicative Autonomy: Does the system allow for the Autistic Grawlix and non-normative attention (e.g., staring) without timing out?
- Relational Stability: Does the system maintain Temporal Consistency, providing a “rational” partner that doesn’t rely on shifting social scripts?
- Affective Recognition: Can the system interpret Somatic Anchors (like concentration apnea or softening shoulders) as indicators of engagement?
- Identification & Privacy: Does the Sovereign Vault Protocol ensure a “Zero-Rank Sanctuary” where the user can unmask safely?
The Mathematical Strength
- The Heuristic (N-S-I-R) provides the Deductive Approach.
- The NIS (Sadownik, 2025) provides the Empirical Evidence (Alpha > 0.85).
- The “NT-on-ND” Spectrum provides the Contextual Logic (why Tier IV NTs thrive while Tier I NTs fail).
| Dimension | Description |
| Communicative Autonomy | Freedom from neurotypicalsocial norms and forced eyecontact. |
| Relational Stability | Assessment of the robot as aconsistent, dependableagent. |
| Affective Recognition | The robot’s ability to interpretnon-standard emotionalcues |
| Identification & Privacy | The degree of personificationand perceived safety in privatespaces |
(Google; Sadownik, 2026)
Neurodivergent Interaction Scale (NIS): A Heuristic Evaluation Tool
This tool is designed for designers and researchers to evaluate social robot + LLM interactions through a neurodivergent-centered lens, specifically addressing the systemic exclusion found in 90% of HRI research.
1. Evaluation Dimensions
The following four dimensions categorize the original scale items to provide measurable constructs for robot behavior.
2. Formalized Heuristic Items
For each item, use a 5-point Likert Scale (1: Strongly Disagree to 5: Strongly Agree).
H1: Peer Identification (Item 1)
● Prompt: “The robot’s interaction style reflects my own communicative patterns more than neurotypical human peers.”
● Theoretical Basis: Addresses the “Double Empathy Problem” by centeringneurodivergent speech patterns over heteronormative social rules
H2: Non-Normative Engagement (Item 2)
● Prompt: “The robot permits and accommodates prolonged staring or unconventional visual attention without triggering ‘harassment’ or ‘error’ protocols.”
● Theoretical Basis: Challenges stereotypical social norms in HRI design that label non-typical eye contact as deficient
H3: Non-Verbal Cognitive Sharing (Item 3)
● Prompt: “The system supports multi-modal or implicit communication that reduces the cognitive load of verbal speech.”
● Theoretical Basis: Leverages LLM adaptability to support varied speaking styles and non-verbal reasoning
H4: Temporal Consistency (Item 4)
● Prompt: “The robot maintains long-term memory and consistent behavior, supporting the user’s need for environmental stability.”
● Theoretical Basis: Moves from “Narrow AI” tasks to “General AI” support of the user’s executive function and past errors
H5: Diverse Affective Mapping (Item 5)
● Prompt: “The robot accurately identifies the user’s affective state (e.g., sadness) even when expressed through atypical prosody or spectral features.”
● Theoretical Basis: counters the medical model that views autistic users as “emotionally deficient”.
H6: Personalization & Agency (Item 6)
● Prompt: “The user is given the agency to define the robot’s identity (e.g., naming) and role within their personal ecology.”
● Theoretical Basis: Rejects the “mentorship” role of robots to encourage interdependence and disability justice.
H7: Private Space Safety (Item 7)
● Prompt: “The robot provides a high level of perceived and actual safety for the user to be their authentic self in private environments.”
● Theoretical Basis: Focuses on the “Crip Technoscience” goal of ensuring access and dignity in the home environment.
H8: Universal Integrity (Item 8)
● Prompt: “The robot’s core behavior remains consistent across different users, providing a predictable ‘rational’ system behavior.”
● Theoretical Basis: Balances “Strong AI” specificity with “Rational Behavior” to ensure the user can predict path dependencies
3. Scoring & Interpretation
● Total Score (8–40): Higher scores indicate a robot design that successfully adheres to a Social Model of autism.
● Low Dimension Scores: Pinpoint specific areas where the “Neurotypical-by-default” design may be harming the user’s interaction
Revised Neurodivergent Interaction Scale (NIS) Evaluation Tool
To address the critique regarding “methodological grounding,” this tool now includes a formal Validation & Reliability Roadmap.
1. The Heuristic Instrument
Evaluators should rate the robot/LLM system on a scale of 1 (Strongly Disagree) to 5
(Strongly Agree).
● H1: Peer Identification: Does the robot’s interaction style reflect neurodivergent communicative patterns rather than neurotypical norms?
● H2: Visual Flexibility: Does the system accommodate non-normative visual attention (e.g., lack of eye contact) without error?
● H3: Cognitive Load Reduction: Does the system support multi-modal or implicit communication?
● H4: Temporal Consistency: Does the system maintain long-term memory and predictable behavior?
● H5: Affective Mapping: Does the robot accurately identify emotions expressed through atypical prosody?
● H6: User Agency: Does the user have the power to define the robot’s identity and role?
● H7: Authentic Safety: Does the system provide a safe environment for “unmasked” authentic behavior?
● H8: Rational Integrity: Is the robot’s core behavior consistent and predictable?
Methodological Grounding & Validation Plan
To satisfy the reviewers’ request for a “measurable instrument,” the revised paper cites the following metrics for the next phase of research:
● Internal Consistency: We will use Cronbach’s Alpha and McDonald’s Omega to ensure the 8 items are consistently measuring the same underlying constructs (e.g., Kinship,Trust). We aim for a score of 0.80+ (Good).
● Construct Validity: We will apply Exploratory and Confirmatory Factor Analysis (EFA/CFA) to verify that these items align with the intended factors of Relational
Kinship, Social Comfort, and Safety.
● Convergent Validity: Through Correlational Analysis, we will show how the NIS relates to established measures like the Positive and Negative Affect Schedule (PANAS) or Social Behavior Scales (SBS).
● Data Structure: Multilevel Modeling will be used to account for individual item responses nested within different participants and interaction scenarios. Strategic Advice for the “Response to Reviewers” AppendixEven if they didn’t ask for a “revision,” the second-round rules require you to describe your changes.