Explore emotional context
Inspect valence, arousal, and dominance through the affective matrix and the one SensAffect eVAD lexicon.
THE EMOTIONAL INTELLIGENCE LAYER
SensAffect exists to help applications respond to the person and the moment. Use emotional context to inform agents, support experiences, and adaptive interfaces—with evidence, user preferences, and clear boundaries guiding the response.
Inspect valence, arousal, and dominance through the affective matrix and the one SensAffect eVAD lexicon.
Explore governance policies and dual authority in the platform playground.
Explore typed contracts, SDK examples, and affective configuration presets.
The same response does not fit every moment. A person exploring, feeling overwhelmed, or asking for help may need different pacing, choices, or support. A single sentiment label leaves out much of the picture. SensAffect represents emotional context across valence, arousal, and dominance, alongside the evidence and policies that inform a response.
How positive or negative is the emotional tone?
Negative PositiveHow activated or settled is the state?
Settled ActivatedHow much control or agency does the state convey?
Overwhelmed In controlInform agents, support experiences, and adaptive interfaces with context that helps them choose a more useful next step.
Give an agent structured context to consider before its next response. Explore when to clarify, slow the exchange, or offer a different path.
Context-aware conversationPrototype responses to frustration and uncertainty. Test policies for when an interaction should move to human support.
Escalation and response policiesExplore affective presets for visual intensity, motion, and interaction pace, with user preferences guiding the experience.
Interface configurationThree application-policy examples show how context can inform a response. Use the source as a starting point for your own integration.
Tailor a shopping experience with quieter layouts, relevant choices, and optional help.
For a seasonal promotion such as a Blue Monday sale, offer a quieter storefront: fewer competing elements, interest-based recommendations, and easy access to help. Keep offers consistent and let the shopper choose their mode.
Context → focused layout and optional assistanceshoppingExperience(context, {
quietMode: true,
interests: ['home', 'music']
});Adapt a music experience to the listener’s goal, from focused listening to winding down.
In a music-streaming app, let the listener choose “unwind.” Suggest a calmer mix, soften visual motion, and lengthen track transitions while keeping their queue and playback controls in their hands.
Listening goal → gentler sound transitions and visualslisteningExperience(context, {
goal: 'unwind',
reducedMotion: false
});Help an agent offer clearer guidance and a timely route to human support.
When the available context suggests frustration, an agent can offer shorter instructions and a route to human support. Evaluate the evidence and offer a choice before taking action.
Context → clearer guidance and a human-support optionsupportExperience(context, {
requestHuman: false
});Runnable JavaScript policy examples with supplied context. These illustrate application decisions; they are not a connected SDK or live emotion detection.
Advanced humanoid implementations could bring the same principles to embodied agents: interpret context, adapt the interaction, and check authority before acting. These are potential applications, not currently available SensAffect robot products.
Support home-care conversations with patient explanations and caregiver coordination.
A humanoid supporting a home-care team could adjust its speaking pace, explain a routine more clearly, or offer to contact a caregiver when someone expresses discomfort or uncertainty.
Emotional context informs communication and requests for help. Clinical decisions, medication, and physical assistance require separate professional oversight and safety controls.
Context → reassurance, clear explanations, caregiver contactAdapt onboard hospitality to each guest’s preferences and conversational pace.
A hospitality humanoid could adapt its conversational style, explain a menu patiently, or offer a quieter interaction when a guest prefers less stimulation.
Guest preferences guide service. Allergens, food handling, alcohol service, and physical movement remain subject to explicit checks; emotional inference does not establish eligibility or consent.
Context → personalised service and human-staff handoffSupport supervised emotional learning through play, storytelling, and conversation.
A friendly humanoid could support age-appropriate emotion vocabulary, turn-taking, storytelling, and practising how to ask a trusted adult for help.
Designed for caregiver-supervised use, with clear disclosure that it is a robot, minimal data collection, and easy stop controls. It should support real-world relationships without encouraging secrecy, dependency, or replacing family, friends, teachers, or professional care.
Context → guided play, emotional learning, trusted-adult supportSensAffect would provide interaction context within a broader robotics system. Physical safety, safeguarding, consent, and domain-specific validation must be designed and evaluated separately.
Extending the humanoid concepts above, proposed capabilities could connect what an agent observes, the objects and people around it, and the continuity of an interaction. The following names describe future scope, not released features.
Explore visual context from scenes, gestures, posture, and movement.
Explore consented visual observations of scenes, posture, gestures, and movement. Combine these signals with conversational context and uncertainty rather than treating appearance as proof of an emotion.
Understand surrounding objects and their spatial relationships.
Represent objects, their spatial relationships, and possible interactions. A care assistant could identify a requested item; a hospitality robot could locate tableware, with separate checks before handling anything.
Connect interactions to trusted people, roles, and authorised handoffs.
Explore a permission-based model of people, roles, and trusted contacts: who is present, whom the user wants involved, and who may authorise or receive a handoff. Recognition alone would not grant access or consent.
Keep a user-controlled record of context, preferences, and follow-ups.
Explore a user-controlled record of interactions, observations, preferences, and follow-ups. Keep observed events distinct from interpretations, with review, correction, deletion, and limited retention built into the experience.
Explore proximity, orientation, personal space, movement trajectories, and changes in the surrounding environment to inform when an agent should approach, pause, or request assistance.
Explore how voice, gaze, posture, gesture, and movement pace can make an agent's intentions easier to understand. Physical actions would remain subject to explicit permission and independent robotics safety controls.
Possible combinations include RealVision and RealObject for home-care or hospitality context, RealContacts for a caregiver or staff handoff, and RealDiary for authorised continuity. Child-facing implementations would require caregiver controls and additional safeguarding across every capability.
A continuous loop connects observations, emotional context, policy evaluation, and outcomes.
Bring observations into a structured representation that your application can work with.
Assemble affective context from evidence and inspect the dimensions behind it.
Check policy and authority before an action is allowed to proceed.
Connect an authorised response to the application and observe what happens next.
How an application uses a signal matters as much as the signal itself. Explore authority, policy, and evidence together in the platform.
Explore the governance playground ↗Use the dual-authority model to distinguish permission to interpret context from permission to take action.
Inspect evaluation results in the governance playground before wiring behaviour into your application.
Work with typed contracts and explicit context instead of treating an inferred emotion as a certainty.
Move from a visual exploration to the contracts and examples used by the platform.
Explore emotional dimensions in an interactive spatial view.
131 curated feeling words — smaller, more appropriate, and more accurate than a general VAD dump. This site uses that lexicon.
Try configuration presets for different interaction styles.
Review structured interfaces and integration examples.
Choose a moment where better context could help. Explore the signal, define the boundaries, and evaluate the response before expanding.
Define the interaction you want to explore and what a useful response would look like.
Use the console to inspect the affective representation and its limitations.
Review the SDK examples, apply your policies, and test against your own requirements.
It is an illustrative multimetric model. Labels cycle through the SensAffect eVAD Lexicon. The homepage does not read your emotions.
Sentiment is one part of the picture. SensAffect also represents valence, arousal, and dominance, with platform tools for exploring evidence, governance, and application behaviour.
Developers and product teams exploring how emotional context can inform agents, support experiences, and adaptive interfaces.
Read the eVAD lexicon on this site, then open the platform for the matrix and governance playground.
No. An affective representation is context for an application to evaluate, not a definitive statement about a person. Evidence, uncertainty, user preferences, and appropriate boundaries still matter.
Give agents, support experiences, and adaptive interfaces the context to respond more thoughtfully. Start with one interaction and evaluate what improves.