Kirsh proposed an alternate model, showing that every stage in Norman's model can have an interactive relationship with the environment. We learn about options by exploring the interface, discover how to specify actions by trying them out and observing the outcomes, position our bodies to better perceive environmental responses, and adjust the environment to facilitate response evaluation.

The cognitive scientist Kirsh presented a criticism of Norman's view of dialogue and developed an alternative based on the theory of embodied cognition [416].

Horvitz [360] summarized the principles of mixed-initiative interfaces as follows:

Communication repair refers to the "work of restoring shared understanding" when conversational partners misunderstand each other [60].

A cornerstone of this research is the book Plans and Situated Action: The Problem of Human–Machine Communication by Suchman [804].

Norman offered two central concepts to help us understand these cognitive efforts: the gulf of execution and the gulf of evaluation (Figure 18.2). These two concepts describe inferential breakpoints for users seeking to express their intentions and interpret feedback from the system, respectively.

Formal models of computation are suitable for describing discrete, moded dialogues. A mode refers to the variation in the interpretation of a user's input according to an internal state. In a modeless dialogue, all inputs are possible in all states and their interpretation is always the same.

Consistency: Are the same actions available, and do they have the same consequences across similar states? Dialogue length: How many turns are needed to get from the initial state to the end state? Number of choices: The number of options available to the user is a predictor of choice reaction time. Error recovery cost: If an error is made, how many turns are needed to recover from it? Connectedness: Can final states be reached from all initial states? Strong connectedness: Can final states be reached from all initial states via a particular action? Reversibility: Can the effect of a given action be reversed in one action?

Dialogue can be described using models of computation from computer science. Such models include finite state machines (FSMs), pushdown automata, and Petri nets. These models can be expressed with formal languages, including context-free grammars and graphs, and they can be implemented in event handlers in user interface (UI) software.

An FSM is a tuple (Σ, S,s0, δ, F), where: • Σ is the input, that is, a finite set of symbols; • S is a finite set of states or modes; • s0 ∈ S is the initial state; • δ is the state transition function δ : S × Σ→S; • F is the set of final states, that is, a subset of S.