The literature regarding trust between a human and a technological system is abundant. In this context, trust does not seem to follow a simple dynamic given the multiple factors that impact it: mode of communication of the system, appearance, severity of possible system failures, factors favoring recovery, etc.In this work, we propose a modeling of the dynamics of the trust of a human agent towards an autonomous system (Human Autonomy Teaming HAT) which is inspired by a hysteresis cycle. The latter reflects a delay in the effect in the behavior of materials called inertia. According to this same principle, the variation in confidence would be based on a non-linear relationship between confidence and expectation. Thus, these variations would appear as interactions occur (like a discrete variable), rather than on a continuous time scale.Furthermore, we suggest that trust varies depending on: the conformity of expectations, the previous level of trust, the duration of maintaining a good or bad level of trust, and the interindividual characteristics of the human agent.Expectations reflect the evaluation of the situation estimated by the human agent on the basis of the knowledge at its disposal and the expected performance of the system. At each confrontation with reality, if the perceived reality agrees with the expected then the expectations are consistent, otherwise they are non-compliant. Depending on the initial state of trust, these expectations will influence the variation in trust. The latter is determined through the hysteresis cycle. At both ends of the cycle, the level of trust is characterized as either calibrated trust or distrust. Indeed, confidence does not increase towards a maximum, but towards an optimal level: calibrated confidence. This is a level of confidence adapted to the capabilities of the autonomous system. Conversely, trust decreases to a level of distrust. This corresponds to the situation where the individual does not trust the system and rejects it. In our context of use, the individual is obliged to continue to interact with the autonomous system, which opens the possibility of overcoming this distrust and restoring all or part of the initial trust.We propose that maintaining this level of calibrated trust or distrust results in an inertia effect. The more trust is maintained at one of these levels, the greater the inertia. Thus, calibrated trust established over a short period of time will be more affected by non-compliant expectations than calibrated trust established over the long term.Furthermore, the evolution of trust is influenced by individual criteria. Although the model described here is generic, it can be personalized according to the predispositions of the human agent: propensity for trust, personality trait, attitudes towards technological systems, etc.The model presented is not intended to debate the nature of trust. It illustrates and explains the dynamics of trust, a key factor in the HAT relationship, both at the origin of this interaction and for the results it produces.

The aim of this paper is to explore the potential challenges associated with contextualizing handover instructions between different stakeholders in the context of a remotely piloted aircraft. The difficulty of synthesizing pilots’ exchanges has already been highlighted. In this study, the operational framework focuses on both the construction of representations and narrative discourse to synthesize the situation. To conduct this study, subjects participated in an experiment in which they were asked to place themselves in a handover situation as an incoming and as an outgoing party. This study has made it possible to identify different strategies for constructing the representation of the situation and for transmitting instructions to contextualize an operation. Using an information gathering analysis method, it is possible to assess not only the relevant elements of the context, but also the narrative progression of these elements in the discourse. This is part of the development of an automated assistant that mediates the sequence and, more generally, the contextualization between two human operators or between a human operator and an artificial agent. Those primary results are the first step towards appropriate standardized protocols and checklists, the implementation of technological solutions such as automation and digital communication systems to improve situational awareness and communication skills. The aim is to identify and apply generic strategies that can be used as a framework for the automated development of effective handover protocols in future UAMs, piloted or remoted-piloted aircraft and other flight operations, contributing to safer and more efficient flight operations.

In the past few years, Reduced Crew Operations (RCO) represent an inevitable as an unavoidable development for the aviation industry, and as such a major challenge for pilots and airlines. It is justified for us, researchers, to address the potential shortcomings of such a development. Over the course of their professional life, pilots might face incapacitation, which may be temporary or permanent, partial or total, coming from cognitive overload, stress, loss of situational awareness (SA), or anything else impairing their ability to carry out their missions. Up to now, the risk of pilots’ occupational incapacity has been addressed by the presence of a copilot in the cockpit, able to take over if necessary. Following this reduction in aircrew, the next step is placing a pilot alone onboard to manage the activities of the crew (case of Single Pilot Operation). It is thus fundamental to design a virtual assistant in order to support the capacity of efficient and relevant knowledge transfer. It is therefore necessary to be able to transfer knowledge of the context to a pilot replacing the one in a state of incapacitation.Since our study is rather explorative, the authors could not cover a whole flight mission. Therefore, this study focused on the landing phase. We placed particularly emphasis in our research on the study of the pilot’s cognitive models such as SA, operational strategy and decision-making. We investigated the inherent behavioral elements to design a decision model adapted to the variability of the pilot’s profiles. Afterwards, we designed a test bench and presented a use case to eighteen pilots from different companies with a representative set of individuality. For this purpose, we placed those pilots in a flight simulation during a complex scenario involving an aircraft failure during a landing phase with bad weather. At the end of the scenario, we asked pilots to go through their experience in the simulator to understand their situation awareness. These situation awareness requirements gave us clues with regard to the pilot’s perception of salient information about the cockpit environment, and the context associated with the situation.This explorative study highlighted the need for understanding how pilots share their knowledge to another pilot. By doing so, we aim to develop a generic methodology of selection of the useful and necessary information to share with an operator in a complex situation.

Considering the ever-growing, deeper interaction between humans and intelligent systems, it is clear that efforts need to be made to ensure a less demanding cognitive workload for the operator. The key to this is bringing the focus onto the Human relationship with machines, ensuring that this interaction is as effective as possible. Any difficulty in understanding at this point can have significant repercussions. Thus, our approach named COMBI, is an interpreter able to enhance communication between humans and Embedded Decision Aid Systems, which can be incredibly complex, to make it mutually intelligible for both parties. This means: (i)Ensuring machines receive in a suitable "language" the operational intention of the Human user; (ii)Clarifying the proposals from the system in a way that can be easily used by humans. In creating a shared and mutually understandable cognitive representation, COMBI contributes to bridging that ontological gap between the high-level abstraction employed by humans and low-level abstraction employed by machines.

In monocular see-through augmented reality systems, each eye is stimulated differently by a monocular image that is superimposed on the binocular background. This can impair binocular fusion, due to interocular conflict. As a function of visual characteristics, the latter can have a greater or lesser impact on user comfort and performance. This study tested several visual characteristics of a binocular background and a monocular element during an exposure that reproduced the interocular conflict induced by a monocular see-through near-eye display. The aim was to identify which factors impact the user's performance. Performance was measured as target tracking and event detection, identification, fixation time, and latency. Our results demonstrate that performance is a function of the binocular background. Furthermore, exogenous attentional stimulation, in the form of a pulse with different levels of contrast applied to the monocular display, appears to preserve performance in most background conditions.

Thales AVS has launched an internal project to develop a solution for in flight data collection and ground tools allowing post-flight analysis of fatigue, sleepiness and sleep of crew members. This solution is developed with support from major airlines. Thirty flights have already been done and collected data show a real interest from air operators to get objective information. This confirms one of the main assumption that subjective measures are not sufficient to have an accurate and reliable evaluation of fatigue, sleepiness and sleep. Such efficient monitoring is a much needed trustworthy tool for Fatigue Risk Management (FRM) and an unavoidable first step towards Single Pilot Operations (SPO), even in cruise. Thanks to these collaborations with air operators, Thales AVS has developed a better understanding of how air operators manage fatigue risks and safety of flight operations and refined its crew monitoring technology for fatigue.

Los pilotos de aviacion civil operan en situaciones dinamicas e inciertas, con la ayuda de sistemas tecnicos complejos. En estos entornos, la gestion de riesgos por parte de los operadores de primera linea debe reducirse al minimo y controlarse continuamente, ya que los pilotos estan sujetos a importantes cargas cognitivas que pueden dar lugar a ciertos errores de pilotaje. Para facilitar la actividad, los disenadores de la cabina de mando y de la interfaz tratan de optimizar la conciencia de situacion y la representacion mental de los pilotos. En este estudio, nos interesan las representaciones mentales de los copilotos durante la realizacion de escenarios de riesgo. En particular, examinamos las diferencias que existen entre la representacion mental «prescrita, esperada» y la «real, efectiva» de los copilotos, destacando los artefactos que estan en la raiz de esas diferencias. Para entender la evolucion de la representacion mental en el curso de una actividad, sometimos a una cohorte de copilotos a un escenario arriesgado. Su actividad fue filmada, comentada por los expertos y luego se realizaron entrevistas de autoconfrontacion para informar los diferentes puntos de vista de la actividad. A traves de la presentacion de un estudio de caso de un copiloto, este articulo propone la descripcion de nuestra herramienta original para la representacion grafica y cronologica de la actividad. Esta herramienta logra destacar en particular las representaciones mentales del copiloto en cada etapa del escenario, las desviaciones que pueden o no haber impactado en la actividad, el uso de las herramientas y el impacto que esto puede tener en las representaciones mentales. Esta herramienta tiene muchas ventajas: una comparacion eficaz de los estados mentales relativos a dos o mas operadores, una presentacion practicamente exhaustiva de los diferentes conjuntos de datos recogidos y analizados, una vision sintetica y precisa de la actividad desplegada. La herramienta de analisis es aplicable a campos distintos de la aeronautica y ofrece un potencial considerable para identificar errores cognitivos y resolverlos en la fuente.

Complexity is everywhere and its progression isn’t weakening. AI, ML, DL, CC, IOT, QC (AI: Artificial Intelligence ∙ ML: Machine Learning ∙ DL: Deep Learning ∙ CC: Constant Connectivity ∙ IOT: Internet of Things ∙ QC: Quantum Computing), … all these barbaric acronyms are already spreading through our daily lives with debatable success. Everyone owns those wonderful fine pieces of equipment (Smartphone, Smart TV, connected appliances, even our basic PCs, … the list goes on and on) that we use without really mastering them. When they perform, we perform (most of the time), but whenever anything goes wrong we become helpless facing a void of incomprehension where we unsurprisingly fail. These technologies can suddenly turn daft, obscure and counter intuitive because their inherent (usually hidden) complexity surface to our interaction. If the situation is critical, consequences can be extremely severe. Pilots can also be in such situation where they have to face the critical emergence of hardly manageable complexity. It’s becoming common in HF related incidents or accidents, where we have the classic: “Pilots didn’t understand what the system was doing and the system never got the pilots intentions”.

Les pilotes de ligne en aviation civile évoluent dans des situations dynamiques et incertaines, assistés par des systèmes techniques complexes. Dans ces environnements, la gestion des risques par les opérateurs de première ligne doit être continuellement minimisée et maîtrisée, les pilotes sont soumis à d’importantes charges cognitives, pouvant mener à certaines erreurs de pilotage. Pour faciliter l’activité, les concepteurs de cockpits et d’interfaces cherchent à optimiser la conscience de situation et la représentation mentale des pilotes. Dans cette étude, nous nous intéressons aux représentations mentales des co-pilotes lors de la réalisation de scénarios risqués. Nous étudions en particulier, les écarts qui existent entre la représentation mentale « prescrite, attendue » et la représentation mentale « réelle, effective » des co-pilotes, soulignant les artéfacts qui sont à l’origine de ces écarts. Pour comprendre l’évolution de la représentation mentale au fil d’une activité, nous avons soumis une cohorte de co-pilotes à la réalisation d’un scénario risqué. L’activité de ces derniers a été filmée, commentée par des experts, puis des entretiens d’auto-confrontation ont été réalisés afin de renseigner les différents points de vue de l’activité. A travers la présentation d’une étude de cas d’un co-pilote, cet article propose la description de notre outil original de représentation graphique et chronologique de l’activité. Cet outil parvient à mettre en avant notamment : les représentations mentales du co-pilote à chaque étape du scénario, les écarts ayant impacté ou non l’activité, l’usage des outils et l’impact que cela peut avoir sur les représentations mentales. Cet outil présente de nombreux avantages : une comparaison efficace d’états mentaux relatifs à deux ou plusieurs opérateurs, une quasi-exhaustivité dans la présentation des différents jeux de données recueillis et analysés, une vision synthétique et précise de l’activité déployée. L’outil d’analyse est applicable à d’autres domaines que l’aéronautique et offre des prises considérables pour identifier les erreurs d’origine cognitive et les solutionner à la source.

Monocular augmented reality devices are used in aviation to help civilian or military pilots in their flying task. In those devices, a virtual image is projected in front of one eye on a see-through glass allowing them to see environment with both eyes. This glass is mounted on a helmet allowing users to maintain the virtual information, usually find in the primary flight display, in front of the eye. Given that the image is presented in front of only one eye, the two eyes are not stimulated in the same way and it can creates a phenomenon known as binocular rivalry. When it appeared, brain is not able to merge the two visuals information and an alternation between them can occur. This alternation appears according to visual condition and the question arises as is it relevant to choose the eye to display the image to limit binocular rivalry and guarantee a good performance in the tasks of recognition and control. The control task consists in matching a monocular item with a binocular item and the detection task consists in giving the right opening orientation of Landolt rings that can be monocular or binocular. Our study aims to compare the performances as a function of the position of the virtual image. These results are then compared to the results of several dominant eye tests to determine if one test can objectively determine on which eye the monocular information should be displayed when using a monocular see-through device.

Monocular augmented reality devices are used in aviation to help civilian or military pilots in their flying task. Given that the image is presented in front of only one eye, the two eyes are not stimulated in the same way and it can create a phenomenon known as binocular rivalry. It appears when the brain is not able to merge the two visuals information and an alternation between them can occur. This alternation is dependent on visual condition and the question arises as is it relevant to choose the eye to display the image to limit binocular rivalry and guarantee a good performance in the tasks of recognition and control. Our study aims to compare the performances as a function of the position of the virtual image. These results are then compared to the results of several dominant eye tests to determine if one test can objectively determine on which eye the monocular information should be displayed when using a monocular see-through device.

EFFECTS OF EXTREME EMERGENCY SITUATION. BEY Christophe Akiani SAS Bordeaux, France HOURLIER Sylvain Thales/ENSC, Human Engineering for Aerospace Laboratory (HEAL) Bordeaux, France. ANDRE Jean-Marc University of Bordeaux, ENSC-BdxINP, IMS UMR CNRS 5218 Bordeaux, France The management of cognitive resources are central in the case of a decision-making process by pilots. We undertake a study involving Airbus 400M pilots and allowing to understand these mechanisms and to propose recommendations for the design of a tool to assist in the management of their cognitive resources. We find that in the most critical cases and under strong temporal pressure, the maintenance of control of the situation corresponds to a survival type behavior which alone can allow a return to the metarules (back to basics). Our display management proposal allows the pilot to maintain control of the situation regardless of his capabilities. It allows a phase of stabilization by a reduction of the stress, then a phase of "soft" recovery of the control and the management of the mission on larger spatiotemporal dimensions. Three modes of entry in the HMI are envisaged: spontaneous, proposed and on demand.

This review aims to clarify the parameters affecting binocular rivalry, in order to improve comfort for users of monocular augmented reality devices. Augmented reality devices allow users to see virtual information superimposed on the environment. The particularity of monocular systems is that they do not stimulate the eyes in the same way and can therefore induce binocular rivalry. This occurs when the brain is unable to merge the different images presented to each eye and perception alternates between them. It can cause visual fatigue, headache and visual suppression. Binocular rivalry can be characterized in terms of alternation rate, predominance (i.e. total proportion of the binocular rivalry viewing time that a stimulus is dominant) and average dominance duration (for all individual dominance periods). The literature suggests that these variables depend on the conditions of use and the visual stimuli available to the subject. Notably, several parameters have an impact, including contrast, spatial frequency, brightness, etc. The impact of other parameters, such as ocular dominance, remains the subject of debate. With respect to the latter, the literature describes various definitions and tests, and it appears that there are three main forms: motor, acuity and sensorial, the latter being of interest for binocular rivalry.

Consumer market touch screens ubiquity has driven the avionics industry to launch in-depth evaluations of touch screens for cockpit integration. This chapter is a follow-up from a 2015 International Symposium on Aviation Psychology (ISAP) paper in which a methodology for turbulence simulation design was presented. One of the challenges was to verify touch screen compatibility with in-flight use under turbulent conditions, ranging from light to severe. The avionics industry recognized early on the need to alleviate such usability risk, and the results of our evaluations enabled us to define recommendations for our HMI designs. Using our validated turbulence profiles, basic touchscreen interaction performances were analyzed, and this paper will focus on the results we gathered using our turbulence simulator.

A monocular augmented reality device allows the user to see information that is superimposed on the environment. As it does not stimulate both eyes in the same way, it creates a phenomenon known as binocular rivalry. The question therefore arises as to whether monocular information should be displayed to a particular eye and if an ocular dominance test can determine it. This paper contributes to give a better understanding of ocular dominance by comparing nine tests. Our results suggest that ocular dominance can be divided into sighting and sensorial dominance. However, different sensorial dominance tests give different results, suggesting that it is composed of distinct components that are assessed by different tests. There is a need for a comprehensive test that can consider all of these components, in order to identify on which eye monocular information should be directed to when using monocular augmented reality devices. Augmented reality refers to an interactive virtual interface combined and real with environment. Within semi-transparent devices), displays can be divided into types: biocular

Adaptive automation/agent has been “a good idea” for 40 years now. Yet it’s hardly used so far. Automations changing in accordance to internal rules are widely distributed and eventually fail to be understandable whenever their inner change can’t be grasped by the operator supposedly “trained & in charge”. All that could change because for technological reasons AI is back with the assumption is that it will fix it all. How can we build cooperative agents capable of helping Humans by building them with a techno-centered view? The epic fail of Ai in the 90’ will just repeat itself. We need to analyze the root of our need when envisioning cooperation with agents. So, what is it that we want from adaptive agents? If you take the example of an assistant surgeon, you have your answer, we want that kind of adaptation. They facilitate the surgeon work without any (verbal) exchanges (not resource demanding to control). They know what to do and when to help. They can interpret any sign from the surgeon as a directive for help. They completely share the same references. They know so well the implicit that collective work seems like the work of a single entity. Alas that is the description of a human being. So definitely what we seek in a cooperative agent are qualities reserved to the living like the ability to adapt. We have misplaced assumptions of humanity on AI without giving it the potential for it: socializing for cooperation through proper communication. It’s called articulation work and it’s been around 30 years at least. It’s the key to enable effective cooperation between agents. DARPA has just realized it and has launched in 2017 a massive research project so as to use AI to digitize the interaction level between an agent and an operator. Modeling with AI what makes a proper cooperation between agents and Human could be the answer.

This report summarizes the discussions and findings of the Workshop on Intelligent Autonomous Agents for Cyber Defence and Resilience organized by the NATO research group IST-152-RTG. The workshop was held in Prague, Czech Republic, on 18-20 October 2017. There is a growing recognition that future cyber defense should involve extensive use of partially autonomous agents that actively patrol the friendly network, and detect and react to hostile activities rapidly (far faster than human reaction time), before the hostile malware is able to inflict major damage, evade friendly agents, or destroy friendly agents. This requires cyber-defense agents with a significant degree of intelligence, autonomy, self-learning, and adaptability. The report focuses on the following questions: In what computing and tactical environments would such an agent operate? What data would be available for the agent to observe or ingest? What actions would the agent be able to take? How would such an agent plan a complex course of actions? Would the agent learn from its experiences, and how? How would the agent collaborate with humans? How can we ensure that the agent will not take undesirable destructive actions? Is it possible to help envision such an agent with a simple example?