Interaction-Readiness in HCI & Beyond

• Interaction-readiness is the condition that enables systems or agents to be interpreted as intentional partners rather than passive data processors.
• It spans disciplines from HCI to mobile health, emphasizing design cues like resistance and legibility to foster goal-directed mutual adjustment.
• Operational models in robotics, social driving, mHealth, and UI automation demonstrate that readiness is dynamically assessed through measurable contextual and behavioral metrics.

Interaction-readiness denotes the conditions under which a system, agent, person, group, or organization is prepared to enter into, sustain, modulate, or terminate interaction in a manner appropriate to the relevant task, medium, and social setting. In a foundational HCI formulation, it is the property of a system that makes it an appropriate partner in interaction among intelligent agents, rather than a passive data processor, and it is tied to “thick” descriptions in terms of goals, intentions, and meaning rather than “just the data” (Müller, 2016). Subsequent work does not converge on a single universal definition. Instead, interaction and related readiness notions are treated as plural, situated, and pragmatically useful across domains such as human–robot interaction, mobile health, phone UI automation, behavior change, driving, ubiquitous computing, and organizational AI adoption (Reeves et al., 2019).

1. Conceptual origin in “new HCI”

In Müller’s account, interaction-readiness belongs to a broader transition from “good old fashioned HCI” to “new HCI,” where HCI becomes “a part of cognitive systems research where HCI is one case of the interaction of intelligent agents” (Müller, 2016). The relevant contrast is between systems treated merely as devices for better data analysis and systems treated, from the human point of view, as intentional partners. Müller ties this to a background claim that “intelligence is the ability to flexibly successfully reach goals,” and therefore that recognizing intelligence requires recognizing that an agent has goals, even if those goals are not fully known (Müller, 2016).

Within this framework, interaction is not exhausted by sending and receiving data. It is goal-directed mutual adjustment between agents each taken to have goals. Interaction-readiness, in this sense, is the system’s disposition to be interpretable as having goals, to display behavior that constrains and is constrained by the human’s goals, and to let the human attribute intentions to it. The paper is explicit that this does not require “fully cognitive agents”; minimal, structured resistance and goal-directed behavior can be sufficient to trigger human attribution of intentions (Müller, 2016).

A central implication is that HCI methodology must move beyond thin descriptions such as signal processing accuracy or syntactic correctness. Müller imports the contrast between thin and thick description to argue that interaction among intelligent agents must be described in terms of meaning, goals, and intentions. This suggests that interaction-readiness is not primarily a property of representational fidelity but of whether a system’s behavior can be thickly described as purposeful in an ongoing interactional context (Müller, 2016).

2. Resistance, intentional attribution, and the legibility of goals

Müller’s most distinctive claim is that intentions are recognized “largely through resistance to carrying out one’s own intentions” (Müller, 2016). Resistance is not treated as a defect in interface design, but as a structural cue that the other entity has its own trajectory, constraints, or goals. Corridor negotiation, steering controls with “feel,” and GUI scrolling with inertia all serve as cases where resistance makes the other side legible as goal-directed rather than inert (Müller, 2016).

The paper’s examples make the design implication precise. It is “easier to negotiate with someone who has their own intentions than with someone who just wants to do whatever I do,” and in a corridor one is “more likely to collide with a super-polite person” than with a “bullish person who clearly indicates where he is going” (Müller, 2016). For HCI, this means that infinite compliance is not equivalent to good interaction. A system becomes interaction-ready when it provides stable, discernible constraints or tendencies that the human can learn and negotiate with.

This extends to embodied and quasi-physical interaction. Pedals, levers, cars, and planes are easier to use when they offer resistance; inertial scrolling is easier to grasp because the list behaves “as though one accelerates an object with a weight” (Müller, 2016). A plausible implication is that interaction-readiness often depends on making system dynamics legible through resistance rather than through explicit explanatory messages alone.

Müller does not supply equations, pseudo-code, or a formal model of resistance; the paper is programmatic and philosophical (Müller, 2016). Even so, it points toward a design orientation in which user action enters a loop with system-side constraints, inertial tendencies, or policy boundaries, and those boundaries are exposed behaviorally. In that sense, interaction-readiness begins once the system’s conduct supports reliable “as-if” intentional attribution.

3. Pluralization of the concept across HCI discourse

Later HCI work argues against fixing “interaction” once and for all. Reeves and Beck bracket the matter as interaction* and treat HCI’s concepts of interaction as “promiscuous concepts”: underdefined, permissive, and generative, capable of coordinating heterogeneous communities without collapsing them into one definition (Reeves et al., 2019). On that view, any use of “interaction-readiness” is most defensible when presented as a situated way of talking rather than as a single universal metric.

This conceptual pragmatics has two consequences. First, interaction-readiness may inherit different meanings depending on whether interaction is framed as dialogue, transmission, control, tool use, embodiment, experience, mediation, or practice (Reeves et al., 2019). Second, attempts to standardize the term too aggressively risk turning it into an “administrative” or disciplinary instrument. Reeves and Beck explicitly argue that underdefinition has been useful for HCI because it supports diverse treatments in practice, and they recommend a therapeutic approach that distinguishes vernacular from technical senses while allowing productive conflict to remain visible (Reeves et al., 2019).

A similar pluralization appears when readiness is tied to behavior change. In mobile physical-activity technology, “readiness” is treated as the stage of behavior change a person is in—precontemplation, contemplation, preparation, action, or maintenance—and this stage is reported as a strong predictor of adoption (Karapanos, 2016). In that paper, readiness is a psychological state that determines what kinds of information, motivation, and support are interactionally appropriate. Individuals in contemplation and preparation had an adoption rate of 56%, whereas those in precontemplation, action, or maintenance had an adoption rate of 20% (Karapanos, 2016). This is not Müller’s concept, but it preserves the same core intuition: successful interaction depends on alignment between system behavior and the user’s current goals, capacities, and constraints.

The combined lesson is that interaction-readiness names a family resemblance across domains. In some literatures it refers to the legibility of a system as an intentional partner; in others it denotes the user’s stage of change, receptivity, or ability to engage; in still others it denotes the semantic, infrastructural, or organizational conditions that make interaction sustainable. This suggests that the term functions best as a controlled plural rather than a singular essence (Reeves et al., 2019).

4. Formal and process-based operationalizations

Several recent frameworks make interaction-readiness computationally explicit. In social and human–robot interaction, IM HERE models engagement through effort, focus, and reciprocal focus relations (Strazdas et al., 3 Dec 2025). An entity’s interpreted effort is defined as

$EI = T \cdot M \cdot C \cdot A \cdot P$

where $T$ is contrast, $M$ magnitude, $C$ contribution, $A$ alignment, and $P$ preference (Strazdas et al., 3 Dec 2025). Focus is assigned to the entity with maximal interpreted effort, and mutual engagement occurs only when two entities have reciprocal focus:

$$Engaged_{A,B} \Leftrightarrow focus_A = B \land focus_B = A.$$

Within this model, readiness is the degree to which an engageable entity is configured to enter into, maintain, or terminate interaction through observable effort and focus modulation (Strazdas et al., 3 Dec 2025).

In social driving, interaction-readiness is formulated as the ability to perceive interaction risk, choose appropriate strategies, and execute temporally coherent responsive actions. A process-based framework computes comprehensive risk from current and predicted motion states, derives benchmark actions from a game-theoretic rational agent model, and scores a real driver’s behavior by similarity to that benchmark (Liu et al., 2024). The final interaction ability score is

$S_i = \frac{Co_i}{1 + D_{MSD,i}} \in [-1,1],$

where $Co_i$ is cosine similarity between observed and benchmark acceleration profiles and $D_{MSD,i}$ is a morphological similarity distance (Liu et al., 2024). Here readiness is not an abstract trait but a dynamic property of the evolving interaction process.

In mHealth, receptivity functions as interaction-readiness for intervention delivery. A just-in-time adaptive intervention system defines receptivity as the person’s ability to receive, process, and use the support provided, and operationalizes a “just-in-time response” as replying to the initiating message within 10 minutes (Mishra et al., 2020). Static and adaptive models based on contextual features improved receptivity relative to random timing by up to 40% in the field deployment, and receptivity to the adaptive model increased over the course of the study (Mishra et al., 2020). The relevant readiness is therefore momentary and inferable from context, rather than assumed.

In phone UI automation, interaction-readiness is formalized as a per-step decision about whether the next action can be performed autonomously or whether user interaction is needed. The task is to detect the need for user interaction and generate an appropriate message from the natural-language instruction, current screen, and action history (Kahlon et al., 25 Mar 2025). AndroidInteraction contains 772 episodes and 3,605 steps, with interaction needed on about 6% of steps, and annotates both binary interaction labels and a 1–5 necessity score (Kahlon et al., 25 Mar 2025). Baseline models remain weak—best F1 is 0.25—which indicates that current LLMs are poor judges of when to interrupt, clarify, or confirm in realistic UI control (Kahlon et al., 25 Mar 2025).

In ubiquitous computing, readiness is reviewed through physiological proxies. Cognitive readiness is treated as a multidimensional construct involving attention, working memory, mental resilience, calculative work, reaction time, errors and accuracy, reasoning, strategic decision-making, and distress tolerance (Chowdhury et al., 7 Jan 2025). For interaction-readiness, these same components can be inferred from HR, HRV, sleep, skin conductance, pupil dilation, eyeblinks, posture, physical activity, EEG, blood oxygen, and hydration, especially in in-the-wild wearable settings (Chowdhury et al., 7 Jan 2025). A plausible implication is that interaction timing, interface complexity, and interruption policy can be personalized by continuously estimated physiological readiness states.

5. Architectural, organizational, and educational extensions

In IoT-type applications, interaction-readiness is treated as an architectural property. The interaction-oriented architecture of networking applications is designed to minimize the effort needed to adapt systems to changing interaction networks, and this readiness is said to be semantic: it depends on how behavior and meaning are modeled at the level of interactions rather than only at the level of computation or transport (Reich, 2017). Systems are interaction-ready when their internal deterministic behavior is cleanly separated from external nondeterministic roles, when protocols and data types define the minimal shared semantics necessary for loose coupling, and when roles can be recombined without widespread rework (Reich, 2017).

At the organizational level, readiness for AI adoption is likewise treated as emergent and interactional. A qualitative study of AI implementation experts argues that organizational readiness develops through “continuous cycles of individual understanding, social learning, and organizational adaptation,” especially as people encounter AI limitations through hands-on use and then share those insights through peers, champions, and governance structures (Übellacker, 21 Feb 2025). This is not readiness for a single interaction episode, but readiness for sustained human–AI interaction in work systems. The study’s core claim is that understanding limitations leads to more realistic expectations and increased trust, and that organizations succeed when they translate these insights into formal workflows and governance (Übellacker, 21 Feb 2025).

A related domain-level perspective appears in firm productivity research. “Domain AI readiness” is defined as “the degree to which an industrial domain is technologically integrated with AI,” measured from IPC4 patent co-occurrence patterns (Zeng et al., 13 Aug 2025). The main performance equation includes an interaction term between firm AI capability and domain readiness, and the paper reports strong complementarity: AI capability yields greater productivity and innovation gains in domains with higher readiness, whereas benefits are limited in domains that are technologically unprepared or already obsolete (Zeng et al., 13 Aug 2025). In this sense, interaction-readiness is the external condition that determines whether firm-level AI capability can successfully “plug into” a domain.

Educational theory extends the notion again. Interactionalism argues that, in the large language agent era, readiness for learning and work means being able to think and act with large language agents rather than alone (Moldoveanu et al., 1 Jan 2025). The paper names this “interactional intelligence,” decomposing it into meta-cognitive operations such as self explicitation, task decomposition, task switching, and objective refinement, and into meta-emotional or meta-relational skills such as inferring user emotionality and designing relationally attuned agents (Moldoveanu et al., 1 Jan 2025). This suggests that interaction-readiness can also be understood as a cultivable human skill set rather than only a property of systems or environments.

6. Group coordination, initiative, and contested design choices

A persistent controversy concerns how much initiative an artificial agent should take. In collaborative human–robot work, a proactive robot that “contributes without being addressed” may offer timely support but may also disrupt coordination, divide attention, or interrupt turn-taking; a reactive robot preserves human control but may miss opportunities to help (Vitry et al., 26 Jun 2026). In a collaborative escape room, a proactive robot that listened continuously, contributed autonomously, and re-initiated interaction after 90 seconds of silence produced substantially more interactions ($T$0, $T$1) than a reactive robot ($T$2, $T$3), yet the reactive condition showed a descriptively higher overall success rate (92.86% vs. 71.42%) (Vitry et al., 26 Jun 2026).

The strongest differences were conditioned by prior experience and personality. Participants with LLM experience solved early puzzles faster in the reactive condition; participants without prior robot or escape-room experience often evaluated proactive behavior more positively; introverted participants showed a mixed pattern in which proactive behavior helped on later, harder puzzles while reactive behavior supported earlier, simpler ones (Vitry et al., 26 Jun 2026). A plausible implication is that interaction-readiness in group settings is not a monotonic variable: being “more ready” to intervene can help or harm depending on user expertise, task phase, and group needs.

A parallel two-dimensional debate appears in the Reality–Artificiality Continuum, which places interaction settings simultaneously on a physical axis from real to virtual and a social axis from real human to artificial companion (Ancona et al., 2020). Reliable interaction across this space requires availability and accessibility, safety, realism and engagement, and ethical compliance (Ancona et al., 2020). The paper proposes ontologies for semantic consistency and runtime verification for physical and social interaction protocols, implying that readiness is partly a matter of being monitorable and correctable at run time rather than merely well designed in advance (Ancona et al., 2020).

These literatures collectively reject a simplistic equation between readiness and immediacy. Sometimes readiness requires resistance rather than compliance, delay rather than promptness, or re-initiation rather than passivity. Sometimes it requires withholding autonomy until a clarification or confirmation has been obtained. The concept therefore sits at the intersection of timing, scope of autonomy, legibility of constraints, and accommodation of user or group differences (Kahlon et al., 25 Mar 2025).

7. Open problems and enduring tensions

Across the surveyed work, one unresolved issue is formalization. Müller’s foundational account is explicit that it provides no equations or explicit formal model of resistance (Müller, 2016). Reeves and Beck, from a different angle, caution that there are reasons interaction concepts may resist formalization and that many interesting concepts can become less useful when defined too precisely (Reeves et al., 2019). By contrast, IM HERE, driving interaction scoring, mHealth receptivity detection, and AndroidInteraction all move toward explicit variables, state models, and evaluation metrics (Strazdas et al., 3 Dec 2025).

A second tension concerns subjectivity and calibration. AndroidInteraction reports only moderate agreement on whether interaction is needed at a given step, with Cohen’s $T$4, and necessity is explicitly treated as varying with user tolerance for interruptions and defaults (Kahlon et al., 25 Mar 2025). In behavior change, stage transitions are reversible and relapse can make confrontation counterproductive (Karapanos, 2016). In organizational AI adoption, realistic trust depends on accumulated experiences with limitations rather than on formal policy alone (Übellacker, 21 Feb 2025). These findings suggest that interaction-readiness is often relational and user-specific rather than a purely objective system state.

A third open problem is scaling from minimal cues to richer cognition. Müller notes a continuum from simple reactive systems such as Braitenberg vehicles to fully cognitive agents, but leaves architectures for scaling up unspecified (Müller, 2016). IM HERE claims generality across human–human, human–robot, and robot–robot interaction, yet also notes limitations in sensing quality, real-world validation, and scalability (Strazdas et al., 3 Dec 2025). Wearable-based cognitive readiness reviews emphasize the complexity of multimodal measurement, individual variability, and the need for a robust catalog of readiness measurements for in-the-wild settings (Chowdhury et al., 7 Jan 2025).

The broad research trajectory therefore remains open. Interaction-readiness can denote a system’s legibility as an intentional partner, a user’s stage of preparedness, an agent’s receptivity estimate, a robot’s calibrated initiative, an architecture’s semantic loose coupling, a domain’s integration with AI, or an organization’s learned ability to work with AI limitations. This plurality does not eliminate the concept’s value. It suggests, instead, that interaction-readiness is best treated as a family of technically precise but context-bound constructs whose common concern is the same: whether interaction at a given moment, with a given partner, under given constraints, can proceed in a way that is interpretable, coordinated, and sustainable (Reeves et al., 2019).