AI in Flight: Advancing Aviation Safety Through Real-Time Monitoring

of Pilots’ Neuropsychological States

L. Gicquel, O. Bartheye and L. Fabre

Centre de Recherche de l’

Ecole de l’Air (CREA),

Ecole de l’Air et de l’Espace, F-13661, Salon-de-Provence, France

Keywords:

Artiﬁcial Intelligence, Aviation Safety, Neuropsychological Monitoring, Real-Time Systems, EEG, Pilot

Cognitive Load, Biometric Sensors, Risk Management.

Abstract:

This study proposes an artiﬁcial intelligence (AI) system to enhance aviation safety by monitoring pilot neu-

ropsychological states in real-time to address human errors in critical situations. Focusing on light aeronautics

as a simpliﬁed model, the system aims to monitor and assess the neuropsychological state of the pilots through-

out ﬂights, in order to assist the pilots in critical situations, offer solutions, and possibly prevent incidents from

happening. Our ﬁrst approach involves the identiﬁcation of cognitive factors and electrophysiological corre-

lates that inﬂuence pilot performance in extreme conditions. The data is extracted from EEG, ECG, and EOG,

combined with camera tracking technologies. This method aims to bridge the current gap between laboratory

research and practical application, ensuring that pilots operate safely even in demanding ﬂight conditions.

1 INTRODUCTION

In the domain of aviation safety, a critical and promis-

ing area of research is the integration of Artiﬁcial

Intelligence (AI) autonomous components with real-

time monitoring of pilot neuropsychological states to

prevent and compensate for human failures in critical

situations.

Numerous studies explore these possibility, in au-

tonomous vehicles research as well as aeronautics.

However, a signiﬁcant gap remains between labora-

tory data, on-terrain recording, and effective use. The

present work is an adaptable proof of concept de-

signed using a simple model of light aeronautics, fo-

cusing on the paraglider, particularly in its competi-

tion (e.g., cross type) or speed variants (i.e., speed-

riding or speed-ﬂying, ﬁgure 1).

The ﬂexible wing pilot is free from many of the

constraints linked to aviation: no engines, less speed,

no complex takeoff checklist, although he also faces

heightened dangers due to the lack of protection pro-

vided by the aircraft, a potential lack of training, and

the limitations of his instrumentation. This makes

speed-ﬂying an interesting prototypal use-case, con-

crete and realistic with fewer complexities to take

into account. Actually, no system has succeeded in

effectively integrating AI agents to proactively man-

age accident situations and ensure the safety of pi-

lots in extreme conditions. This is the aim of our

study, which deﬁnes a generic and adaptive concept

for exploiting the capabilities of AI to warn and guide

pilots throughout the ﬂight and to assist or replace

them in the event of a ﬂight failure or incident. To

achieve these ends, we detailed a method structured

around two key actions: 1) Identify the factors that

can inﬂuence cognition in an ‘extreme’ use case and

the electrophysiological correlates based on our pre-

vious studies (Mass

e et al., 2022; Melani et al., 2023;

Melani et al., ) and 2) Propose an AI algorithm tai-

lored for aeronautics (Lee, 2006) and, a risk diagram

and a conceptual analysis of a monitoring system dur-

ing pilots’ activity.

Previous works showed that emotions, cognitive

fatigue or mental work load are factors to be taken

into account in aviation safety (Dehais et al., 2018;

onmez and Uslu, 2018; Holtzer et al., 2010; Marcus

and Rosekind, 2017; Mass

e et al., 2022; Melani et al.,

2023; Mass

e et al., 2022). For example, Mass

e et

al. (2022) emphasized the importance of monitoring

pilot cognitive workload and cognitive fatigue to pre-

vent inattentional deafness and enhance ﬂight safety.

In their experiment, participants had to detect rare

sounds in an ecological context of simulated ﬂight

under cognitive fatigue. Mass

e et al. (2022) found

that participants performed better to detect alarms un-

der low cognitive load conditions compared with high

cognitive load conditions. Results showed that alarm

omission and alarm detection can be classiﬁed us-

Gicquel, L., Bartheye, O. and Fabre, L.

AI in Flight: Advancing Aviation Safety Through Real-Time Monitoring of Pilots’ Neuropsychological States.

DOI: 10.5220/0012977800004562

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 2nd International Conference on Cognitive Aircraft Systems (ICCAS 2024), pages 80-87

ISBN: 978-989-758-724-5

ing explainable AI and machine learning computation

based on time-frequency analysis of brain activity.

Melani et al. (under review) asked participants to

verify complex multiplication problems that were ei-

ther true (e.g., 3×23=69) or false (e.g., 5×98=485).

They found that negative emotions modulated some,

but not all, arithmetic problem veriﬁcation mecha-

nisms, with electrophysiological variations observed

across different problem types. Interestingly, Melani

et al. found that negative emotions inﬂuence problem

processing from early stages, as evidenced by modu-

lations in early ERP components such as P1 and N2

but also other components such as the P300, the P600

and LPC. In summary, previous studies have shown

that cognitive fatigue, mental load and emotions can

modify the way in which individuals solve problems.

Thus, under certain conditions, participants may

not use the right problem-solving strategies, leading

to a drop in performance. These factors not only in-

ﬂuence cognitive performance, but may also be as-

sociated with changes in electrophysiological compo-

nents. Consequently, it is essential to take these fac-

tors into account and identify them in order to main-

tain air safety. This is what we do by determining the

electrophysical tools that can be used for this partic-

ular use case. A comprehensive literature study al-

lowed to outline the physiological characteristics and

metrics that can be used for classiﬁcation and make

the connection with the BCI.

The system would record these inputs using a

combination of portable and non-invasive methods,

ensuring minimal interference with the operator’s

tasks. Data from EEG, ECG, and PPG are pro-

cessed using advanced algorithms such as wavelet

transforms and machine learning models, including

SVM, to accurately classify different psychological

and physiological states. Eye and camera tracking

technologies provide additional data points for assess-

ing vigilance and cognitive load through ﬁxation met-

rics, saccades, and pupillary responses.

2 METHODS

The ﬂexible wing pilot is, ﬁrst and foremost, a “ﬂyer,”

and the basic principles of ﬂight remain the same. The

psychological and physical risks incurred are largely

similar, and a psychological or physical failure could

lead to a serious incident if it is not taken care of by

the pilot himself or by external assistance.

2.1 Determining the Inﬂuencing Factors

Increasing Risks to the Soft-Wing

Pilot in Flight

To explore factors inﬂuencing pilots’ cognition and

decision-making, especially in critical situations, we

conducted semi-structured interviews with ten soft

wing pilots (min age = 25 years, max age = 60; min

ﬂight experience = 50 h, max ﬂight experience = 2000

h). Each interview session lasted between one and

two hours, with informal exchange periods.

The pilots’ experience ranged from former test

pilots to paragliding and speed-ﬂying instructors, as

well as leisure pilots with variable experience (be-

tween 2 to 15 years). This diverse sample allowed

us to examine accident factors and how these factors

vary with pilot experience and ﬂight conditions. In

this initial phase of our research, we chose not to em-

ploy a structured coding strategy, as our main aim was

to explore the breadth and depth of pilots’ cognitive

experiences and decision-making processes in a way

that remained open to the rich narratives provided by

participants.

This approach allowed us to capture the com-

plexity of pilot experiences without constraining their

responses to predeﬁned categories. This narrative

method enabled us to identify key themes and pat-

terns that may not have been apparent with a more

restrictive coding scheme. Interviews were designed

to elicit accounts of routine as well as emergency

decision-making processes; pilots recounted memo-

rable ﬂights and incidents as well as ﬂight incident

simulation practice detailing their state of mind, ac-

tion timing and reactions. The questions focused on

a three-paneled timeline, before, during and after the

ﬂight, to characterize the psychological situations that

the pilot undergoes throughout the duration of each

phase.

Looking forward, as our research progresses and

the initial exploratory ﬁndings are further reﬁned, we

anticipate utilizing qualitative data analysis software

such as NVivo to perform more structured thematic

analysis. This will enable us to systematically cate-

gorize and analyze the data, providing a means to val-

idate and extend our preliminary insights. Employ-

ing NVivo will also facilitate a rigorous examination

of the relationships between themes and sub-themes

and, help quantify the prevalence of speciﬁc experi-

ences and viewpoints, thereby enhancing the robust-

ness and generalizability of our results.

AI in Flight: Advancing Aviation Safety Through Real-Time Monitoring of Pilots’ Neuropsychological States

2.2 Electrophysiological Monitoring of

the Psychological Factors

To be able to analyze the general situational aware-

ness of the pilot, multiple biometrics sensors can be

reviewed. The inputs to the system should include a

variety of physiological and behavioral data sources:

• EEG (Electroencephalography): For monitor-

ing brain activities, which help assess levels of

attention and cognitive load. The mental state

is evaluated thanks to EEG micro-states. Brain

wave monitoring also allows for speciﬁc drowsi-

ness and emotion change detection.

• ECG (Electrocardiography): Used to track

heart rate variability, which can indicate stress

levels.

• PPG (Photoplethysmography): For measuring

pulse rate variability, providing insights into the

pilots’ cardiovascular health and stress responses.

• Eye Tracking / Camera Tracking: These tech-

nologies are critical for evaluating cognitive load

and attention by analyzing eye movements, blink

rates, and pupil dilation. This helps in understand-

ing how pilots focus and react under different ﬂy-

ing conditions.

• EOG (Electro-oculography) Recording: This

captures eye movement artifacts in addition to

EEG data, further enhancing the detection of fa-

tigue and cognitive load.

• Physiological Indicators: Measurements of skin

conductance and body temperature help in assess-

ing stress levels through changes in the autonomic

nervous system.

Most of these wearable sensors are miniaturized.

For the EEG recording, we chose to use a dry EEG

with 4 electrodes.

2.3 Adaptation of the Design Method to

Speedﬂying

The soft wing paradigm also allows to model quite a

variety of ﬂights, as it covers a few types of wings.

A very small wing (8-10sqm) will allow for high

speed and proximity ﬂying risk modeling, a bigger

one (16-19sqm) will allow for longer ﬂights and give

more time to the system to propose another outcome

than reserve launching. Bigger wings like competitive

paragliders can go up to thousands of meters of alti-

tude and allow for very long ﬂights (up to ten hours),

which gives space and time to test more complicated

scenarios and fatigue or drowsiness onset and subse-

quent recordings.

3 RESULTS

The initial interview protocol facilitated the identiﬁ-

cation of various risk factors pertinent to speed-ﬂying,

encompassing cognitive load, mental stress, emo-

tional state, drowsiness, and fatigue. Cognitive load

denotes the equilibrium between task demands and

available time for completion, initially inferred from

experimental ﬁndings related to problem-solving and

learning outcomes, gauged through performance met-

rics and subjective inquiries. Pilots expressed its im-

pact as ”having too much on one’s mind,” attribut-

ing it to personal and professional concerns that di-

minish attention span and focus capacity. Evaluation

of cognitive load may involve EEG or eye-tracking

methodologies. Mental stress, partly intertwined with

cognitive load, denotes a psychological state arising

when individuals perceive demands surpassing cop-

ing abilities, instigating psychological and physiolog-

ical ﬁght-or-ﬂight responses. Participants reported

mental stress beyond cognitive load issues, notably

in proximity ﬂying, acrobatic maneuvers, or extended

ﬂights, often exacerbating during instances of high

cognitive load and negative emotions, adversely af-

fecting ﬂight quality and occasionally leading to ac-

cidents. Mental stress can precipitate vigilance lapses

or misdirected focus. Drowsiness commonly surfaces

during extended ﬂights, associated with early waking

or insufﬁcient sleep, particularly prevalent in compet-

itive settings. Fatigue emerges post-repeated ﬂights

or intense thermal ﬂights, often correlated with larger

wing usage, albeit beyond the scope of this prelimi-

nary study.

Proposing electrophysiological tools within a BCI

system is imperative for assessing and quantifying

these factors speciﬁc to speed-ﬂying. The BCI sys-

tem should continuously and unobtrusively monitor

pilots’ neuropsychological and physiological states

during ﬂight, correlating physiological risks with cog-

nitive alterations. Extreme environmental condi-

tions and physiological anomalies during ﬂight sig-

niﬁcantly impact pilot psychological states. Criti-

cal physiological episodes affecting pilots encompass

oxygen deprivation, acceleration effects, spatial dis-

orientation, exposure to toxins, and various physio-

logical stresses induced by diverse ﬂying conditions,

exacerbated by hypoxia at high altitudes. Additional

risks include hypocapnia and hypercapnia affecting

cerebral blood ﬂow, extreme gravitational forces lead-

ing to G-LOC, and environmental factors like vi-

brations, temperature ﬂuctuations, and hydration lev-

els. Even minimal instances of these conditions,

not reaching pathological levels, could impair pilot

decision-making, necessitating consideration.

ICCAS 2024 - International Conference on Cognitive Aircraft Systems

Figure 1: Speedﬂying.

3.1 Prototypal Scenario and

Risk-Management Strategies

To establish the various scenarios, we conducted ex-

tensive research on available accident reports from the

French free ﬂight federation (FFVL), videos of ﬂight

incidents, and third-party reports, in order to be able

to create realistic use-case diagrams. The proposed

prototypal scenario is a ﬂight with an obstacle cross-

ing the ﬂight path.

It is to be noted that the soft-wing pilot is freed

from many of the constraints associated with aviation:

no engines, less speed, no complex take-off checklist.

He also faces exacerbated dangers due to the lack of

protection provided by the aircraft, a potential lack

of training, and the limitations of his instrumentation.

The project intends to incorporate risk management

strategies from cybersecurity, adapted from methods

like EBIOS-RM or CORAS, to address the diverse

risks associated with aviation, as well as the system’s

security itself and the speciﬁc risks linked to AI use

in safety systems. The CORAS method ((Den Braber

et al., 2006), (Stolen et al., 2002)) addresses security-

critical systems in general, but places particular em-

phasis on IT security. An IT system for CORAS is not

just technology, but rather a medium for communica-

tion and interaction between different groups of stake-

holders involved in a risk assessment; what matters

is the human interacting with the technology and all

relevant aspects of the surrounding organization and

society (Table 1).

Risks are events that harm assets when they occur.

However, often some risks are accepted, either be-

cause of shortage of resources or because of conﬂict-

ing concerns.

Scales for the likelihood of which incidents occur

are deﬁned (certain, likely, possible, unlikely, rare) in-

cluding the impact or consequence they have on the

assets. Assets are ranked according to their impor-

tance to distinguish risks that can be accepted from

those that cannot.

During the requirements phase, designers and pi-

lots establish the need for a system and document its

purpose. Security planning should begin at the re-

quirements phase and consist of activities to establish

security requirements and assess security needs. The

security risk assessment helps us to deﬁne the func-

tional characteristics of the system requiring a secu-

rity need. During this stage, teams also determine the

need for threat modeling, reviews, and security de-

sign. At the requirements stage, availability and in-

tegrity of system services are taken into account as

well as privacy and data sensitivity measures. Re-

quirement diagrams (see ﬁgure 2) deﬁne the technical

(hardware and software) requirements of a system ca-

pable of real-time monitoring of pilots’ neurological

and physiological state. BPMN diagrams (ﬁgure 3)

represent the various ﬂight scenarios where the sys-

tem should monitor the pilot biometrics, and when it

should intervene or interact with the pilot.

4 DISCUSSION

The present concept is based on multiple data inputs.

Considering there is a need for simpliﬁcation of the

data stream to allow for real-time processing, we in-

tend to implement camera eye tracking and EOG anal-

ysis to classify the cognitive load of the pilot on ﬁrst

intention, as EEG data could be primarily used for the

detection of drowsiness, fatigue or stress.

The cognitive load analysis involves eye tracking

AI in Flight: Advancing Aviation Safety Through Real-Time Monitoring of Pilots’ Neuropsychological States

Table 1: IT system.

Initiator Incident Cause Risk Treatment

(Pilot) Obstacle not seen;

can cause a collision

Drowsiness / Deﬁ-

cient mental state

Collision Allow AI device to

take control

(AI) Emotional state not

detected (false neg-

ative); cannot detect

deﬁcient mental state

Inefﬁcient machine

learning model or

incomplete dataset

Unrecovered human

failure

Improve model and

dataset

to monitor where the pilot directs his gaze, and corre-

lation to EEG data to analyse his information process-

ing. Vigilance and attention can thus be assessed by a

combination of EEG and eye tracking (EOG record-

ing and/or camera tracking). Consequently, our con-

cept aims to quantify precisely speciﬁc eye move-

ment and qualify an increase in blinking and slower

eye movements as signs of fatigue. Constant eye fo-

cus and a lack of blinking would, conversely, indicate

high levels of vigilance and attention.

While it is possible to quantify blinking using

cameras only, the modularity of this system could al-

low to maintain the quality of the assistance when

the use of cameras is impossible. To account for

physiological changes and to support this data for

neurophysiological state assessment, we intend to

complement this setting with GSR (skin conductance

and temperature), ECG (heart rate and variability),

and possibly plethysmography data, which measures

changes in volume in different parts of the body.

4.1 Computational Data Analysis

On the computational level, according to Al-Shargie

and al.(Al-shargie et al., 2016), as EEG signals

are non-stationary, wavelet transform can provide

clear features that are strongly correlated with lev-

els of mental stress. Since this study, several other

teams have successfully used Support Vector Machine

(SVM) algorithms and other machine learning meth-

ods to characterize stress, generally achieving detec-

tion levels around 95%, especially when combining

multiple models. Another possibility could be to use

non-supervised methods.

An interesting work on this subject describes a po-

tential contribution of artiﬁcial intelligence and deep

learning (Lee et al., 2023) to the ﬂight environment.

The study presents an autonomous system for EEG-

based multiple abnormal mental states classiﬁcation

using hybrid deep neural networks. Various speciﬁc

abnormal mental states (namely, low fatigue, high fa-

tigue, low workload, high workload, low distraction,

and high distraction) are classiﬁed by applying the

deep learning method. The accuracy is, as customary

with deep learning networks, quite high, but a second

line of classifying, based on traditional algorithmic

rules, could be applied to make the output of the deep

learning model robust and secure in critical cases.

4.2 System Outputs and Performance

The ﬁrst model we have chosen for the sake of sim-

plicity is the mid-range wing, around 16 sqm. Even if

in some cases (thermal ﬂights) the aircraft could the-

oretically have the altitude and remaining ﬂight time

for the system to propose various solutions (calling an

operator, taking some time to wake the pilot), speed-

ﬂying accidents usually go fast.

As such, the preferred output of the system here

would be, as soon as it is detected that the pilot is

unresponsive, to ﬁre a pyrotechnic parachute system

(which have very recently been introduced to the soft

wing world). More complicated outputs will be as-

sessed by modelling incidents with bigger aircrafts

and/or longer ﬂights. We could then establish a con-

nection between an external operator or a discussion

between a speciﬁcally trained large language model

(LLM) to guide the pilot to landing or keep him aware

enough to solve the situation.

The outputs of the system should provide a com-

prehensive assessment of the operator’s readiness and

condition. These outputs include:

• State Detection: Identifying levels of cognitive

load, stress, and attention states, which allows for

timely interventions to prevent accidents or errors.

• Risk Prediction: Forecasting potential physio-

psychological risks such as extreme fatigue or

high stress, which could impair performance.

• Adaptive Feedback: Offering real-time, adaptive

feedback to the operator or automated systems on

board, which can adjust the workload or provide

alerts to mitigate risks.

ICCAS 2024 - International Conference on Cognitive Aircraft Systems

Figure 2: Requirement diagram of the speedﬂying system.

In the speedﬂying use case, often, the best action

of the system would be to prevent the ﬂight scenario

from occurring entirely. A public report for a deadly

accident in 2022 from FFVL presented this as a con-

clusion. One of the inﬂuencing factors in the incident

was that “The pilot had not had a good night’s sleep

and was reluctant to go ﬂying.” These kind of situa-

tions should be taken into account by a general health

assessment by the biometrics sensors as well as some

routine questions before ﬂight, which could be formu-

lated in a natural speech style by a trained LLM.

In case of a strongly dangerous situation, the sys-

tem should aim to prevent the pilot from ﬂying by

informing him and, possibly, calling a person that

would be pre-designated as a safeguard, to try and

discourage the pilot from ﬂying on this particular in-

stance. This system thus acts as a critical tool in en-

hancing the operational safety and efﬁciency of pilots

and other aerospace operators by providing a real-

time, integrated view of their psycho-physiological

status and adapting to their immediate needs.

The risk-management method adapted here to a

speciﬁc prototypal use-case, speedﬂying, is extensi-

ble to any type of aeronautical use-case, by adding

new actors, taking into account a larger array of in-

ﬂuencing factors as well as the speciﬁcs of the con-

sidered aircraft and types of ﬂight. Speciﬁc physio-

logical factors linked to extreme altitude, life support

systems or pressurization of the cabin could be added

in the assessment. Fatigue and drowsiness, which

are generally linked to longer ﬂights, could also be

monitored with the same sensor setup. For example,

drowsiness can be detected with a high accuracy by

EEG (Balandong et al., 2018). The recording area of

preference would, in this case, be the occipital cortex;

we could dedicate a limited number of electrodes for

this task.

In conclusion, this BCI system designed from the

knowledge collected is of real interest in improving

the safety of paragliding ﬂights and is intended to be

generalized to other forms of ﬂight (Deng and others,

2020). The groundwork is set for the proposal of a

similar device that would function on a more compli-

cated use-case, military aviation. Future efforts will

concentrate on developing and training AI algorithms

to process and interpret complex data, as well as gain-

ing pilot acceptance of this kind of products. The cre-

ation of a ﬁrst simpliﬁed prototype based on this re-

search would be possible to this effect.

AI in Flight: Advancing Aviation Safety Through Real-Time Monitoring of Pilots’ Neuropsychological States

Figure 3: BPMN diagram of a ﬂight with an obstacle scenario.

REFERENCES

Al-shargie, F. M., Tang, T. B., Badruddin, N., and Kiguchi,

M. (2016). Mental Stress Quantiﬁcation Using EEG

Signals. In Ibrahim, F., Usman, J., Mohktar, M. S.,

and Ahmad (

Eds, M. Y., editors, International Con-

ference for Innovation in Biomedical Engineering and

Life Sciences, pages 15–19. Springer.

Balandong, R. P., Ahmad, R. F., N., M. S. M., and Malik,

A. S. (2018). A Review on EEG-Based Automatic

Sleepiness Detection Systems for Driver. IEEE Ac-

cess, 6:22908–22919.

Dehais, F., Dupres, A., Di Flumeri, G., Verdiere, K., Borgh-

ini, G., Babiloni, F., and Roy, R. (2018). Monitoring

ICCAS 2024 - International Conference on Cognitive Aircraft Systems

pilot’s cognitive fatigue with engagement features in

simulated and actual ﬂight conditions using an hybrid

fNIRS-EEG passive BCI. In 2018 IEEE international

conference on systems, man, and cybernetics (SMC),

pages 544–549. IEEE.

Den Braber, F., Brændeland, G., Dahl, H. E., Engan, I.,

Hogganvik, I., Lund, M. S., others, and Vraalsen, F.

(2006). The CORAS model-based method for secu-

rity risk analysis. SINTEF, Oslo, 12:15–32.

Deng, P.-Y. and others (2020). Detecting fatigue status of

pilots based on deep learning network using EEG sig-

nals. IEEE Transactions on Cognitive and Develop-

mental Systems, 13(3):575–585.

onmez, K. and Uslu, S. (2018). The relationship be-

tween ﬂight operations and organizations in aircraft

accidents; the application of the human factor analysis

and classiﬁcation system. Anadolu University Journal

of Science and Technology A - Applied Sciences and

Engineering, pages 1–1.

Holtzer, R., Shuman, M., Mahoney, J. R., Lipton, R., and

Verghese, J. (2010). Cognitive Fatigue Deﬁned in the

Context of Attention Networks. Aging, Neuropsychol-

ogy, and Cognition, 18(1):108–128.

Lee, D. H., Jeong, J. H., Yu, B. W., Kam, T. E., and Lee,

S. W. (2023). Autonomous system for EEG-based mul-

tiple abnormal mental states classiﬁcation using hy-

brid deep neural networks under ﬂight environment.

Systems, IEEE Transactions on Systems, Man, and

Cybernetics.

Lee, W.-K. (2006). Risk assessment modeling in aviation

safety management. Journal of Air Transport Man-

agement, 12(5):267–73.

Marcus, J. H. and Rosekind, M. R. (2017). Fatigue in trans-

portation: NTSB investigations and safety recommen-

dations. Injury prevention, 23(4):232–238. Publisher:

BMJ Publishing Group Ltd.

Mass

e, E., Bartheye, O., and Fabre, L. (2022). Classiﬁca-

tion of electrophysiological signatures with explain-

able artiﬁcial intelligence : The case of alarm detec-

tion in ﬂight simulator. Frontiers in Neuroinformatics,

16.

Melani, P., Fabre, L., and Lemaire, P. How negative emo-

tions inﬂuence arithmetic performance : a compari-

son of young and older adults.

Melani, P., Fabre, L., and Lemaire, P. (2023). How neg-

ative emotions inﬂuence arithmetic performance : a

comparison of young and older adults. Current Psy-

chology.

Stolen, K., Den Braber, F., Dimitrakos, T., Fredriksen, R.,

Gran, B. A., Houmb, S. H., others, and Aagedal, J. O.

(2002). Model-based risk assessment–the CORAS ap-

proach. In iTrust Workshop.

AI in Flight: Advancing Aviation Safety Through Real-Time Monitoring of Pilots’ Neuropsychological States