Other terms
AI systems
The notion of an “intelligent system” is neither defined precisely nor demarcated sharply from other systems, artifacts, or technical devices. Instead, the perception of what is judged to be intelligent changes with progress and exposure to such a system. Broadly speaking, an intelligent system is commonly understood as a computational system – such as a search engine, an online shopping assistant, a chat bot, or a cleaning robot – that leverages concepts, tools, and techniques from artificial intelligence in order to establish capabilities that are commonly attributed to humans while (still) being less typical of other soft- and hardware systems. Most notably, these capabilities let an AI system learn from experience and adapt to specific environmental conditions. As a consequence, an intelligent system exhibits a certain degree of autonomy, and its behavior is not completely prespecified. Importantly, intelligent systems are able to interact with humans or other systems through various modalities, for example, textual, visual, acoustic, or haptic signals. Whereas we share contemporary definitions of intelligent systems, we specifically focus on the abilities of AI systems to learn not only from prespecified data but also through interacting with humans.
Co-construction
Co-construction refers to an interactive and iterative process of negotiating both the explanandum and the form of understanding for explanations. It is a process that is performed mutually by sequentially building on, refining, and modifying the interaction: Each partner elaborates upon the other partner’s last contribution. The processes of scaffolding and monitoring (see below) guide this elaboration and direct it toward a specific form of understanding. In effect, what is achieved is the participation of both partners in moving toward a goal. Whereas the co-construction of an explanation takes place on the microlevel of an unfolding interaction and can thus be accessed directly, the process is modulated crucially on the macrolevel of the interaction.
Explainee
The addressee of an explanation
Explainer
The person who steers an explanation forward
Explanandum
The entity (event, phenomenon) that is the subject of an explanation
Explanans
The (verbal) way that an explanation can be expressed and co-constructed by both partners
Monitoring
In this multimodal process, the observed outcome is compared to what was predicted (Pickering and Garrod 2013). Via monitoring, the partners multimodally (i.e., using speech, gestures and nonverbal behavior) keep track of the progress in a joint task. For example, the explainer will monitor the explainee’s understanding by evaluating whether her or his way of explaining has been successful or whether further elaboration or modification is needed. Vice versa, the explainee will monitor the explainer by accepting a level of detail that is needed for a particular explanation.
Scaffolding
In developmental literature, scaffolding refers to the way an expert provides guidance to a learner within a learning process by increasing or reducing the level of assistance in accordance with the partner’s performance. In our approach, we transfer the term from the area of learning to understanding. In accordance with the idea that understanding is constructed by both partners, both partners can scaffold each other—that is, provide the other partner with the information needed to arrive at a joint construction of the explanandum and the desired form of understanding. Together with the process of monitoring, it is not only a form of guidance but also supervision, and both together aid the active participation of both partners.
Understanding
Whereas in the current debate on explainable systems (XAI), understanding refers to the problem of receiving “enough information” (Miller 2019, p. 11), in our approach, understanding is linked to what is relevant for the explainee. To account for variations in the progress of and the varying goals of explanations, we will differentiate between practices of enabling and comprehension. With enabling, we refer to explanations in the context of choosing or performing an action. Comprehension, in contrast, accounts for a reflexive awareness that may lead to a conceptual framework for a phenomenon that goes beyond what is immediately perceivable. We expect further differentiations that will be explored in the individual projects.
Social-practice
Social practice determines the social relations and power structures in a given situation and thereby provides a specific (normative) background for the way interaction will play out in order to ‘place’ the explanation appropriately, and finally how that explanation will be interpreted. Social practice is a product of our actions with respect to each other that often has both social consequences and social presuppositions. The consequences on the one hand and the presuppositions on the other hand speak to the two timescales that constitute a social practice: In terms of consequences, every explaining process re-establishes the relevant social practice; in terms of presuppositions, in turn, the experience of an explaining process will confirm or make a new contribution to our expectations, roles, and partner models in relation to this particular social practice.
%-------------------------- Project A01 ----------------------------------------------
dyad
In sociology, a dyad (from the Greek: δυάς dyás, "pair") is a group of two people, the smallest possible social group. As an adjective, "dyadic" describes their interaction. The pair of individuals in a dyad can be linked via romantic interest, family relation, interests, work, partners in crime, and so on. The relation can be based on equality, but may be based on an asymmetrical or hierarchical relationship (master–servant).
partner-model
a partner model is a main resource is for ‘placing’ explanations and contains knowledge and assumptions about the explainee with regard to her/his dialogical role, general characteristics, or even this specific person
Obligation
Obligations represent what an agent should do, according to some set of norms. The notion of obligation has been studied for many centuries, and its formal aspects are examined using Deontic Logic.
obligor
one who is bound by a legal obligation
obligee
one to whom another is obligated (as by a contract). specifically : one who is protected by a surety bond
Audience-design
Audience design is a process in a symmetric interaction by which speakers tailor what they say in order for the addressee to understand it. Critically, audience design involves taking into account a representation of the addressee’s perspective, and how it differs from one’s own perspective.
interlocutor
one who takes part in dialogue or conversation
reappraisal
re-interpreting or re-analyzing the emotional situation and/or goals
Persuasion
Persuasion can be seen as a further strategy to achieve a decision or behavior that is congruent with logical argumentation and not influenced by emotional processes.
feedback signals
Feedback signals are generally (i) short (i.e., consist of minimal verbal/vocal expressions), (ii) locally adapted to their prosodic context (i.e., the speaker’s utterance) by being more similar in pitch to their immediate surrounding than regular utterances, or (iii) taking place in the visual modality, for example as head gestures or facial expressions.
verbal-feedback
we consider feedback ‘verbal/vocal’, if it is spoken, i.e., produced as a speech sound in the vocal tract of a listener. Examples of such feedback found in the alico-corpus are genau (‘exactly’), ja (‘yes’), mhm (‘uh-huh’), and m.
explanation-purpose
Explanations are provided to support transparency, where users can see some aspects of the inner state or functionality of the AI system. When AI is used as a decision aid, users would seek to use explanations to improve their decision making. If the system behaved unexpectedly or erroneously, users would want explanations for scrutability and debugging to be able to identify the offending fault and take control to make corrections. Indeed, this goal is important and has been well studied regarding user models and debugging intelligent agents. Finally, explanations are often proposed to improve trust in the system and specifically moderate trust to an appropriate level.
transparency
The level to which a system provides information about its internal workings or structure, and the data it has been trained with – this is similar to Lipton’s definition of transparency
fact
that what happened
foil
that what is expected or plausible to happen
causal-explanation
refers to an explanation that is focused on selected causes relevant to interpreting the observation with respect to existing knowledge.
EXPLAINING-WHY
It is a semantic type of explanation which explicates how a complex matter comes into being (e.g., explaining natural phenomena by reference to physical principles, or explaining a person’s action by explicating possible motives.
EXPLAINING-HOW
It is a semantic type of explanation which outlines procedural knowledge about processes and coordinations of actions in order to achieve a specific goal.
EXPLAINING-WHAT
It is a semantic type of explanation which describes, for example, the meaning of a term or a proverb. We consider these distinctions to be useful for describing ways of explaining technical artifacts because they reflect their intrinsic duality.
dialog-act
In linguistics and in particular in natural language understanding, a dialog act is an utterance, in the context of a conversational dialog, that serves a function in the dialog. Types of dialog acts include a question, a statement, or a request for action. Dialog acts are a type of speech act.
%----------------------------------------- Project A03 % reference: https://dash.harvard.edu/bitstream/handle/1/37143006/Feelings-and-Consumer-Decision-Making.pdf;jsessionid=0B25B0CD7E2DADFDE111D056759F3020?sequence=1
%The Appraisal Tendency Framework (ATF) (Lerner, Han, et al. 2007) posits a close relationship between human information processing, decision making, and emotional status. ATF is based on research showing that feeling in a certain way results in specific cognitive processing—the way that information is attended, encoded, stored, and retrieved (Duncan and Barrett 2007). This affects, for example, how explanations are construed (Collins 1996) and possibly processed (Gasper and Clore 2002; Schwarz and Clore 2003).
%Appraisal Dimensions
\section{Appraisal-Dimensions}
any aspect of the criteria that account for a person's evaluation of an interaction with the environment and the generation of an emotional response following the appraisal. Examples of studied appraisal dimensions include the goal relevance of an event, its stimulus novelty, its pleasantness or unpleasantness, and an individual judgment of one
%Appraisal themes %Appraisal tendencies
\section{Appraisal}
An appraisal is a cognitive representation which represents an evaluation of the relevance of some triggering object or event to the organism. (Definition from: Emotion Ontology)
\section{Appraisal theory} Appraisal theory is the theory in psychology that emotions are extracted from our evaluations (appraisals or estimates) of events that cause specific reactions in different people. Essentially, our appraisal of a situation causes an emotional, or affective, response that is going to be based on that appraisal.An example of this is going on a first date. If the date is perceived as positive, one might feel happiness, joy, giddiness, excitement, and/or anticipation, because they have appraised this event as one that could have positive long-term effects, i.e. starting a new relationship, engagement, or even marriage. On the other hand, if the date is perceived negatively, then our emotions, as a result, might include dejection, sadness, emptiness, or fear. Reasoning and understanding of one's emotional reaction becomes important for future appraisals as well. The important aspect of the appraisal theory is that it accounts for individual variability in emotional reactions to the same event.
% emotion regulation (ER) strategies in cognitive behavioral therapy, in which the goal is to learn to change undesired emotions such as anxiety through cognitive strategies.
%cognitive bias
\section{emotion-regulation}
Emotion regulation is the ability to exert control over one’s own emotional state. It may involve behaviors such as rethinking a challenging situation to reduce anger or anxiety, hiding visible signs of sadness or fear, or focusing on reasons to feel happy or calm.
%pleasure, arousal, and dominance
%reference: https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=9093955 %Persuation %Compliance Gaining Behaviors (CGBs)
%virtual agent %continuous guidance of visual attention %micro level factor %task level factor %Visual Focus of Attention (VFOA) %cognitive overload %inattention %visual short-term memory capacity, K, %overall visual capacity C, %attentional weights for individual objects, w, and %selectivity between target and distractor objects, α. %the temporal-order judgment (TOJ)
\section{Pragmatic frames} Pragmatic frames are multimodal interaction patterns that are performed with a partner. They can be characterized by actions that the partners need to fulfill together in order to achieve a joint goal. Crucial to this concept is the idea that when repeating actions for a specific task, partners develop an awareness for the task’s goal.
% -------------------------------------------
\section{Causal attribution} Causal attribution refers to the articulation of internal or external factors that could be attributed to influence the outcome or observation. Miller argues that this is not strictly a causal explanation, since it does not precisely identify key causes. Nevertheless, they provide broad information from which users can judge and identify potential causes. Combining attribution across time and sequence would lead to a causal chain, which is considered a trace explanation or line of reasoning.
% \newglossaryentry{contrastive explanation}{% % name={contrastive-explanation},% % description={}}
% \newglossaryentry{counterfactual explanation}{% % name={counterfactual explanationn},% % description={}}
% \newglossaryentry{transfactual reasoning}{% % name={transfactual reasoning},% % description={}}
% XAI Elements. We identify building blocks that compose many XAI explanations. By identifying these elements, we can determine if an explanation strategy has covered information that could provide key or useful information to users. This reveals how some explanations are just reformulations of the same explanation types but with different representations, such that the information provided and interpretability may be similar. Currently popular is showing feature attribution or influence by indicating which input feature of a model is important or whether it had positive or negative influence towards an outcome. Another intuitive approach highlights similar (or different) instances from the training data, prototypes, criticism, and counter-examples. Other elements include the name and value of input or outputs (generally shown by default in explanations, but fundamental to transparency), and the clause to describe if the value of a feature is above or below a threshold (i.e., a rule).
%
\section{social act} Reinach uses the term ‘social act’ for what in the Anglo-Saxon tradition are called ‘speech acts’, both terms referring in effect to the same sets of instances, namely to those cases where human beings use language (understood as including moans, cries, gestures, and so forth, including gestures involving pointing to lines on a map) in social interactions. Both Searle and Reinach took particular interest in what might be called normative social acts such as acts of promising, whose performance gives rise immediately to a claim and obligation on the part of promiser and promisee (Mulligan 1987).Reinach defines a social act more precisely as an act per- formed by a human being to bring about some end by influencing another human being, typically by using sentences containing verbs such as ‘promise’ or ‘order’ or ‘request’. A social act may involve direct interaction between one person and another, but many social acts achieve their ends indirectly through use of documents. Legal enactments7 are one example of social acts of this sort.
% -------------------- Project B02 ------------------
%reference: https://journals.sagepub.com/doi/pdf/10.1177/1461445616683596 %social practices on three levels: communicative jobs, devices, and forms. %the knowledge gap (the explanandum) %embodied ‘moves’ (devices and forms), %the type of explanandum (more or less accessible) %displays of understanding
% \newglossaryentry{embodied moves}{% % name={embodied-moves},% % description={}}
\section{Jobs} Jobs are global communicative tasks to be accomplished jointly by the participants in Dyad/interactional team. Jobs represent organizational tasks that have to be interactively fulfilled by the participants in the joint achievement of a discourse unit. In accordance with Wald, discourse units are defined in terms of their unity, brought about – among other features – by their being delimited from the surrounding turn-by-turn talk. Since the organizing jobs explicate the steps that are derived from the particular genre’s communicative function, the jobs vary according to the respective genre. Due to their genre- orientation, the jobs are assumed to be applicable for all instances of the particular genre and are thus context-free: if a discourse unit was successfully performed, these tasks must have been fulfilled, no matter how or by whom.
\section{Devices}
moves of participants in the discourse unit (e.g. principal speaker – recipient) which are designed to the fulfillment of the organizational jobs and the requirements of a particular genre. As opposed to jobs, the devices address each participant’s moves in their pragmatic and semantic function with regard to the overarching interactional jobs of the discourse unit. In other words, ‘devices’ address the single, mostly ‘sentence-level’ contributions made by each participant in the joint endeavor to tackle the jobs. However, by virtue of their constituting the complex unit, devices in this framework are not ‘recognizable actions’ by themselves but derive their particular quality as building blocks from the overarching discourse unit and its genre.
\section{Forms}
Forms are linguistic, prosodic and nonverbal resources for realizing the devices. The forms refer to the linguistic and embodied resources for realizing the functional role of the particular device. They constitute the unit’s audible and visible production and can be seen analytically as the level on which participants mark structures and functions and thus help the co-participant to identify the sort of discourse unit being underway. Achieving recognizability is done by forms. As devices and forms are different for different genres as well as for the different interactional roles, specific forms for establishing topical relevance or dissent are also resources for projecting the particular type of discourse unit: ‘This morning I had a row’ is heard as initiating a story (see excerpt 1) due to the reference to a tellable event; ‘Why are you making the dumpling wet?’ (see excerpt 2) states the lack of relevant knowledge and thus asks for an explanation; and ‘But they also need fully qualified grooms’ (see excerpt 3) contradicts an assumed claim and thus opens argumentation (see ‘Interactional jobs across genres and across contexts’ later for a necessarily short sample analysis of closing-devices and forms).
\section{genres} genres can be seen as routinized communicative solutions to recurrent communicative problems of a speech community
\section{discourse unit} A discourse unit is an identifiable structure within a discourse. The form of a discourse unit is partly determined by genre and partly by its social context.
\section{Discourse}
Discourse is the name given to stretches of language longer than one sentence, e.g., conversations, narratives, arguments, and speeches.
%Moves such as rephrasing, introducing extraneous information (Chi, Siler, Jeong, et al. 2001), or metacognitive monitoring through metacommunication (Roscoe and Chi 2008) in this sense have been identified and defined as strategies or pedagogical behaviors applied in tutoring situations (Merrill et al. 1995; Chi, Siler, Jeong, et al. 2001). %
%2.2.6a. Learning from hearing tutors’ explanations. As one can see from Fig. 2, the prominent activity undertaken by the tutor is giving extensive explanations: over half of the tutors’talk consisted of giving explanations. There are two intriguing questions. First, why did tutor explanations correlate with learning at all when other studies have shown that students do not seem to learn from either tutors’ explanations (VanLehn et al., in press) or from listening to a partner’s explanations in a collaborative context (Webb, 1989)? Second, why did tutor explanations correlate with shallow learning only? An explanation of text materials in the context of this tutoring task can be construed to be “helpful” or “useful” for learning in six ways. First, an explanation can rephrase the text sentences well, perhaps by linking the content of the current sentence to prior knowledge introduced earlier in the text, as well as integrating it with information to-be-introduced in the upcoming sentences. Second, an explanation can enhance learning if it contained extraneous information that can fill gaps in the text sentences, or provide a more complete and coherent understanding. Third, an explanation can reinforce learning if it repeats the information in the text. Fourth, an explanation can promote learning if it is rephrased in more understandable everyday language (such as in a more conversational style using nontechnical jargon), rather than in the expository style of a written text. Fifth, tutor explanations can enhance learning if they elicited constructive responses from the students. Finally, from an instructional point of view, an explanation is considered a good one if it is given in response to students’ direct queries or when students signal that they don’t understand (Leinhardt, in press). Presumably, we would like “good instructional explanations” to be targeted at the students’ confusion, lack of understanding, and misunderstanding.
% \newglossaryentry{metacognitive monitoring}{%
% name={metacognitive-monitoring},%
% description={}}
% types of clarification????? % \newglossaryentry{asking for clarification}{% % name={asking for clarification},% % description={is a type of explanation move}}
% \newglossaryentry{providing correct linguistic forms}{% % name={providing correct linguistic forms},% % description={is a type of explanation move}}
% \newglossaryentry{connecting to prior knowledge}{% % name={connecting to prior knowledge},% % description={is a type of explanation move}}
% \newglossaryentry{maintaining understanding throughout an explanation}{% % name={maintaining understanding throughout an explanation},% % description={is a type of explanation move}}
%Child: I runned super fast. %Adult: wow, you ran super fast? %Child: Yeah, I ran super fast.
%Within this framework, social roles can be described as dialogical roles that are constituted by employing specific, role-associated devices and embodied forms to accomplish the respective conversational job.
% \newglossaryentry{dialogical role}{% % name={dialogical-role},% % description={}}
%------------------------------------------ Done
%reference: https://journals.sagepub.com/doi/pdf/10.3102/00028312031001104 %Content of information sought. Table 1 presents the 18 question-content categories in the GPH scheme (Graesser, Person et al., 1992). These categories are defined according to the content of the information sought rather than on the question stems (i.e., why, how, where, etc.) For example, antecedent ques- tions tap previous events and states that cause or enable an event to occur. Antecedent questions can be articulated linguistically with a variety of stems: "Why did the event occur?" "How did the event occur?" "What caused the event to occur?" and so on. Most of the questions have an interrogative syn- tactic form.
\newglossaryentry{question} a question is defined as a speech act that is either an inquiry (as defined previously), an interrogative expression, that is, an utterance that would be followed by a question mark in print, or both. The following expressions are inquiries, but only the first is an interrogative expression: "What is a factorial design?" (interrogative mood), "Tell me what a factorial design is" (imperative mood), and "I need to know what a factorial design is" (declarative mood). All of the following expressions are in the interrogative mood, but only the first is an inquiry: "What is a factorial design?" (inquiry), "Could you stop the session in 5 minutes?" (indirect request), and "Why did I ever take this course?" (gripe).
%reference: https://link.springer.com/content/pdf/10.1007/s11251-007-9034-5.pdf
%Moves such as rephrasing, introducing extraneous information
%or metacognitive monitoring through metacommunication (Roscoe and Chi 2008)
% \newglossaryentry{metacommunication}{% % name={metacommunication},% % description={}}
%reference: https://journals.sagepub.com/doi/pdf/10.1177/1461445616683596
%reference: https://link.springer.com/content/pdf/10.1007/s11251-007-9034-5.pdf
%scaffolding episodes
% \newglossaryentry{scaffolding episode}{%
% name={scaffolding-episode},%
% description={}}
%new dialogical roles such as information collector and information analyzer % \newglossaryentry{information collector}{% % name={information collector},% % description={is a type of dialogical role where...}}
% \newglossaryentry{information analyzer}{% % name={information analyzer},% % description={is a type of dialogical role where...}}
%different question types and reformulations, such as asking for clarification (Clarke et al. 2017), placing information in common ground (Clark and Bernicot 2008), providing correct linguistic forms (Demetras et al. 1986), supporting children’s ongoing narrative (Quasthoff and Kern 2007), connecting to prior knowledge (Rohlfing 2011; Grimminger, Rohlfing, Lüke, et al. 2020), or maintaining understanding throughout an explanation (Kern 2020) can be functional in a dialogue.
\section{metacognitive process}
metacognitive processes such as comprehension-monitoring and metamemory, which involve evaluation of the quality of one’s own knowledge and understanding
% \newglossaryentry{comprehension-monitoring}{% % name={comprehension-monitoring},% % description={}}
%knowledge-building episodes
%behavioral moves (devices and forms) %less accessible explanandum
%metacommunication as a move (apart from questions): we furthermore derive the hypothesis that applying metacommunication as a move (apart from questions) can be particularly valuable for less accessible explananda that are likely to require more monitoring of and agreement on understanding from both partners.
%iconic gestures %beat gestures %prosodic marking %observer viewpoint in gestures %character viewpoints in gestures
%----------------------------------------------------- Done
%users, roles, and explanations in real-world contexts
%Interpretable to Whom? A Role-based Model for Analyzing Interpretable Machine Learning Systems: https://arxiv.org/pdf/1806.07552.pdf %system creators, operators, executers, decision subjects, data subjects, system examiners
\section{machine learning ecosystem} includes the system and the agents that have interactions with, or are affected by, this system. An ecosystem always contains just one machine learning system and one or more agents (in the real-world, ecosystems will often overlap).
\section{Interpretability} The level to which an agent gains, and can make use of, both the information embedded within explanations given by the system and the information provided by the system’s transparency level.
\section{system creators} Agents that create the machine learning system. Several teams of creators may work on different aspects of the same system e.g., architecture, design, implementation, training, documentation, deployment, and maintenance.
\section{Operators} Agents that interact directly with the machine learning system. Operators provide the system with inputs, and directly receive the system’s outputs. In some cases they may be able to interact directly with the system’s creators. Ecosystems always contain operators.
\section{Executors} agents who make decisions that are in- formed by the machine learning system. Executors receive information from operators. Ecosystems al- ways contain executors.
\section{Decision-subjects} agents who are affected by decision(s) made by the executor(s).
\section{Data-subjects} agents whose personal data has been used to train the machine learning system. Ecosystems only contain data-subjects if the machine learning system has been trained on personal data.
\section{Examiners} agents auditing or investigating the ma- chine learning system. Depending on the system, they may interact with one or more of the other roles and the machine learning system itself. Ecosystems only contain examiners when the system is being audited/inspected.
%------------------------------------------------------- Done
%Stakeholders in Explainable AI: https://arxiv.org/pdf/1810.00184.pdf
%In our own recent work, we examined explainability and interpretability from the perspective of explanation recipients, of six kinds (Tomsett et al. 2018): system creators, system operators, executors making a decision on the basis of system outputs, decision subjects affected by an executor’s decision, data subjects whose personal data is used to train a system, and system examiners, e.g., auditors or ombudsmen. We found this Interpretable to whom? framework useful in thinking about what constitutes an acceptable explanation or interpretation for each type of recipient. In this paper, we take a slightly different tack, examining the stakeholder communities around explainable AI, and arguing that there are useful distinctions to be made between stakeholders’ motivations, which lead to further refinement of the classical AI distinction between developers and end-users.
%Preece et al. (2018) conceptualized four AI-related stakeholder communities (developers, theorists, ethicists, users) for which they assume different intents and requirements regarding explainability.
%developers, theorists, ethicists, users
\section{developers} people concerned with building AI applications. Many members of this community are in industry — large corporates and small/medium enterprises — or the public sector, though some are academics or researchers creating systems for a variety of reasons including to assist them with their work.
\section{theorists} people concerned with understanding and advancing AI theory, particularly around deep neural networks. Members of this community tend to be in academic or industrial research units. Many are also active practitioners, though the theorist community is distinguished from developers by their chief motivation being to advance the state of the art in AI rather than deliver practical applications.
\section{ethicists} people concerned with fairness, accountability and transparency1 of AI systems, including policy-makers, commentators, and critics. While this community includes many computer scientists and engineers, it is widely interdisciplinary, including social scientists, lawyers, journalists, economists, and politicians.
\section{users} people who use AI systems
%-------------------------------------------------------------- %Project B04: Explaining as a sociotechnically contextualized normative practice %----> Discussion with Philipp on the norms and the ontological fertility
%----------- Project B05: Co-constructing explainability with an interactively learning robot
%mental model (i.e., internal representations that humans build about things)
%With respect to measuring mental models about robots, in their recent review, Wallkotter et al. (2020) identified three main categories for the measurement of explainability: \section{Trust} measures via a self-report scale how willing a user is to agree with a decision or a plan of a robot and how confident a user is about the robot’s internal functioning.
\section{Robustness} Robustness is most often an observational measure that is concerned with failures during interaction (counting how often a goal is achieved)
\section{Efficiency} Efficiency measures how quickly the task is completed (e.g., in human–robot teams)
%Project B05: This project aims at measuring explainability and works with the assumption that a person is only able to explain a system correctly, when it has been understood: the better a system is understood, the better the explanation can be. We thus aim to measure understanding in users. Understanding in the view of the TRR is linked to what is relevant for the human user, which, in the CCT setting, is the learning process (i.e., how the robot learns). %
%-------------------------------------- % TODO: a list of cognitive processes to be added
%distrustability %distrust %healthy (rather than blind) distrust—that is, a distrust that is specific and founded in %observations.
%laypersons= \section{layperson} a person without professional or specialized knowledge in a particular subject.
%\newglossaryentry{range–frequency-bias}{% % name={range–frequency-bias},% % description={a bias that assigns less probability to the categories judged most likely %and more probability to the other categories}}
%-------------------------------------------- %Project C01 seems to be not much relevant to the Explanations Ontology
%--------------------------------------------
%https://www.researchgate.net/publication/32229632_Simulating_Human_Tutor_Dialog_Moves_in_AutoTutor
%We found that normal human tutors prefer dialog moves that are carefully tailored to the previous student contribution. More specifically, human tutors choose dialog moves that are sensitive to the quality and quantity of the preceding student turn. The tutor dialog move categories that we identified in human tutoring sessions are provided below.
%(1) Positive immediate feedback. "That’s right" "Yeah"
%(2) Neutral immediate feedback. "Okay" "Uh-huh"
%(3) Negative immediate feedback. "Not quite" "No"
%(4) Pumping for more information. "Uh-huh" "What else"
%((5) Prompting for specific information. "The primary memories of the CPU are
%(ROM and ..."
%(%((6) Hinting. "The hard disk can be used for storage" or “What about the hard
%(disk?”
%((7) Elaborating. “CD ROM is another storage medium.”
%((8) Splicing in/correcting content after a student error.
%((9) Summarizing. "So to recap," <succinct recap of answer to question>
% \newglossaryentry{Address statement}{% % name={EM-Address-statement},% % description={Any utterance that addresses a preceding statement, assert or reassert, without being an explicit reject, reject part, accept or accept part.}}
\section{Active student learning} Instead of the student being a passive recipient of information, the educational experience should encourage active student learning.
\section{Sophisticated pedagogical strategies} A good teacher/tutor should implement sophisticated pedagogical strategies that are effective in promoting learning.
\section{Anchored learning in specific examples and cases} A good teacher/tutor should anchor the material in specific examples and cases rather than relying on didactic, declarative information.
\section{Collaborative problem solving and question answering} A good learning experience involves a balanced collaboration between the teacher/tutor and the student while they solve problems and answer questions.
\section{Convergence toward shared meanings} The teacher/student should achieve shared knowledge, a ‘meeting of the minds’.
%---------------------------------------------------- %https://ocw.mit.edu/courses/comparative-media-studies-writing/21w-732-science-writing-and-new-media-fall-2010/readings/MIT21W_732F10_listening.pdf
% MIT Sloan Communication Program Teaching Note % by JoAnne Yates, Sloan Distinguished Professor of Management
% Active Listening and Reflective Responses % One of the basic building blocks of communication--and one of the most difficult skills to learn and practice--is effective listening. We all spend much of our time hearing other people speak, but not necessarily listening to what they are really saying. Instead, for example, we daydream, start formulating responses before hearing the whole point, and interrupt to make a tangential comment. Meanwhile, we have failed to understand the speaker's real point and the thoughts or feelings underlying that point. % As a consultant or manager, you will be called on to listen in many different contexts and for many purposes. You will need to gather data about a problem in order to help solve it. You will need to listen to a subordinate's career problems in order to help him develop. You will need to understand the point of view of another person in order to carry on effective negotiations for something you want from her. You will need to interview candidates for positions in your group. Before you can communicate effectively as managers you need to learn some useful approaches and techniques for effective listening. This teaching note describes active listening, a comprehensive approach to the task of listening. It also describes reflective responses, a particular responding technique that is based upon Western concepts of the role of feelings in interpersonal relationships. Reflective responses can be especially useful in certain types of listening situations. % Active Listening % Active listening is a term often used to describe a general approach to listening that helps you gain more information, improve your understanding of other points of view, and work cooperatively with superiors, subordinates, and peers. This approach requires not just that you learn and remember more of what the other party has said, but also that you communicate your interest and involvement to that party, as well. Active listening requires effective use of verbal and nonverbal communication, as well as mental and emotional discipline. % An active listener: % Looks and sounds interested in the speaker. By conveying your interest, you can encourage the speaker to communicate more extensively and to clarify and expand on thoughts and feelings. Communicate your interest by maintaining good eye contact. (In American culture and many Western cultures, this means looking into the other person's eyes much, though not all, of the time. Too much eye contact may make the speaker feel self-conscious, but too little will make him feel ignored). Maintain a body position and facial expression that indicate attentiveness, not boredom. Nod encouragingly to show understanding and interest. Avoid drawing, playing with your pen, or other distracting behaviors. (Conversely, try not to be distracted by the speaker's mannerisms.) Use vocalizations such as "uh-huh" and "yes" to encourage him to continue. % Adopts the speaker's point of view. You will understand and remember the speaker's points most effectively if you try to see things from her point of view, at least initially. Try to listen, not to interrupt, finish sentences, or rush the speaker. Most of all, try to suppress your initial reactions and to hear and understand the speaker's perspective. Try to listen and respond from the speaker's frame of reference, not your own. Listen for her feelings, not just her words. Try to empathize with her position. Depending on the context and purpose of your communication, you may later shift modes into a discussion in which you also present your own point of view, but to be a good active listener, you shouldn't do that until you thoroughly understand the speaker's point of view. % Clarifies the speaker's thoughts and feelings. You will listen better if you are not talking too much yourself. When you are in active listening mode, limit your talking to things that will contribute to getting the fullest informational and emotional content from the speaker. Avoid inserting your own marginally related experiences and minimize interruptions. When the speaker pauses, ask open-ended questions (e.g., "How do you feel about X?" "Tell me about X." "What concerns you about X?") rather than questions that can be answered in a single word or phrase ("Are you satisfied with X?" "Is X on schedule?'). Use reflective response techniques (described in more detail below) to check the accuracy of your understanding of the speaker's ideas and especial1y feelings (e.g., "So you are frustrated at your inability to progress on X?"). You may adapt active listening techniques to different types of listening, with your role ranging from minimal intervention to more active solicitation and even into giving % advice, depending on your needs. % Reflective Responses % Reflection, or reflective response technique, borrowed from certain types of counseling techniques, is designed to elicit as full a sense as possible of the speaker's thoughts and especially feelings. It is a way of helping someone explore her own personal meanings. This technique involves reflecting back to the speaker what you believe she has said in order to verify (or clarify) your understanding and to encourage the speaker to continue elaborating on her point of view.* [*This description of the technique is based on Chapters 9 and 10 of Interpersonal Behavior: Communication and Understanding in Relationships, by Anthony G. Athos and John J. Gabarro (Prentice Hall)]. An active listener is already using % aspects of this technique, but reflection requires taking even greater care in the following areas: % Reflect the speaker's thoughts and feelings. Restate what you believe the speaker has said to check for the accuracy of your understanding (e.g., "So you couldn't finish the assignment on time." "Then you think the time allotted was inadequate?"). Even more importantly, reflect back the speaker's feelings as you have heard or inferred them (e.g., "You seem to feel anxious because you couldn't finish the assignment on time."). This interpretation of feelings is, of course, more tricky in that it often requires you to read between the lines, to infer feelings underlying what has been said (e.g., "You seem angry about the reorganization," rather than "So the department was reorganized."). Thus you may want to use wording or voice tone to make your inferences into questions, rather than statements (e.g., "So you feel anxious because you couldn't finish the assignment on time?" "Do you feel anxious because you couldn't finish the assignment on time?"). % Respond rather than lead the conversation. Let the speaker's thoughts' and feelings be your guide in the conversation. Don't guide the conversation by asking questions or interjecting ideas or suggestions that take the speaker into new areas of interest to you (e.g., "Have you thought about X?" "Maybe you should ask about Y."). Instead, respond to and reflect back what he actually said or what you sense is implied by what he said (e.g., "So you feel trapped by this project?"). Try to stay within the speaker's frame of reference, rather than asking questions or making suggestions that come from your own frame of reference. Even asking "Why?" may distract the speaker from one line of thoughts or feelings into defending and justifying feelings or actions. % Respond to feelings, rather than content. As suggested earlier, feelings are generally a better indication of personal meanings than content is. Thus you will help the speaker's self-exploration more by responding to her feelings (e.g., "So you resent Susan's frequent absences?") than to the content (e.g., "How often was Susan out of the office?") There is a corollary to this guideline: You can get at a person's feelings better by responding to the more personal aspects of what she says. Whenever possible, choose the specific, personal points (e.g., "So your subordinate, John, tends to challenge you in front of others.") rather than the abstract generalizations (e.g., "So the firm's hierarchy is weak ") to respond to and reflect back at the speaker. This encourages her to explore those personal feelings more thoroughly and to make clear her own assumptions.
% Reflective vs. Directive Responses % Obviously, this reflective technique is not always appropriate to the circumstances and to your needs or purposes. At times you may want to be more directive and less reflective in your interactions. You may want to argue, advise, or confront. Thus once you have learned to use the reflective mode of listening, you need to consider when to use it, when to shift from that mode to a more directive mode, and when not to be reflective at all. Here are some reasons and times for using this reflective technique: % • When you need or want to understand the other person's feelings more completely % • When you sense that the other person has not yet revealed his thoughts and feelings about % the situation % • When you sense that the other person is not sure of his true feelings % Thus active listening with reflective responses is often the first stage of an interaction. Then, once you feel you really understand the person's perspective, you can switch to a more directive or confrontational or persuasive stance. Here, you can lead as well as respond and speak from your own frame of reference as well as the other person's. % Conclusion % Listening is a critical communication skill for managers and consultants, as well as for all of us in our personal lives. Advising someone well on a career, personal, or organizational issue requires that you understand that person's point of view. You can't negotiate effectively until you understand what the other person wants. Effective persuasion depends on a clear understanding of the other person's perspective. In all of these situations, active listening, often beginning with reflective responses, is crucial to achieving your ultimate communication objectives. Active listening is a skill that, like other communication skills, must be developed. It does not come naturally to most of us. Practicing active listening and particularly reflective responses can feel artificial when these skills are isolated from a real communication need. Still, only by practicing can you develop these skills and then integrate them with your other communication skills.
%-------------------------------------------------------------------