Methodology for dialogue management in human- machine systems under the risks

Abstract

This paper analyzes the features of the concept of " Operator 5.0 ." and considers the problem of dialogue session management in complex human-machine systems under conditions of changing environment and psychophysiological state of the operator. The concept of agent approach to dialogue management is proposed, requirements to agent-manager for dialogue management are developed. Models allowing to manage the process of dialogue "man-machine" in conditions of risk and to adapt the course of dialogue to the current conditions of the operator's activity are developed. Examples from various subject areas (the process of controlling an unmanned aerial vehicle, dialogic interaction in an e-learning system) are given to demonstrate the practical implementation of the proposed approach.

Keywords

• eol>Ergonomics
• human-machine interaction
• reliability
• optimization
• interface
• Operator 5
• 0 1

1. Introduction

In modern conditions, when the control of complex technological systems (such as power grids, aviation complexes, industrial robots) becomes more and more automated, the risks associated with the human factor increase [1-4]. The cost of operator error can be catastrophic [5-7]. Traditional systems of human-machine interaction often proceed from the assumption of stability of the external environment and constancy of the operator's psychophysiological state [ 8-10 ]. In reality, these factors are subject to significant changes.

Situations of fatigue, stress, as well as external influences such as a continuous flow of tasks, noise, vibration or equipment malfunctions can significantly reduce the efficiency and reliability of control. Added to this are new threats, in particular cyber-attacks that can compromise the integrity of data and control signals. There is an urgent need to develop adaptive human-machine systems that can dynamically adjust the dialogue session depending on the current situation to prevent errors and ensure security [10-12].

2. Analysis of the research area and problem statement

In recent years, more and more attention has been paid to the problem of human factors: conceptual issues of ergonomics [ 13-17 ], human-computer interaction [ 17-19 ], harmful environmental factors [ 20-21 ], human-operator reliability and risks [ 22-24 ] The task of constructing adaptive human-machine interfaces [ 25-29 ] is particularly emphasized Before the Second World War, the concept of human-machine interaction was based on the principle “humans adapt to machines”, whereas after the war, research in engineering psychology and ergonomics came to the conclusion that “machines must adapt to humans” [ 30 ]. This often proved to be a difficult task, especially in the case of inherently maladaptive systems (the changing complexity and pace of work had a catastrophic effect on the state of the operator). The next step in the development of the methodology was cognitive systems in which two-way adaptive interaction between human and machine was possible. For many years the human-centered approach became the basis of ergonomics. Today ergonomics is moving from a human-centered concept to a humancentered method using artificial intelligence [ 13-14, 29 ], the concept of this approach is generalized in Fig. 1.

The Operator 5.0 concept represents the evolution of the human worker in the era of Industry 5.0, emphasizing a human-centric approach that uses technology for the well-being and health of the operator, not just to improve productivity. In this model, ergonomics is critical to designing systems that take into account a person's physical, cognitive and sensory capabilities, using wearable sensors, motion capture technologies and machine learning to objectively assess risks, monitor fatigue and optimize workplaces to reduce musculoskeletal disorders and improve overall working conditions.

The main principles of the Operator 5.0 concept are [ 13-14, 29 ]:   

Human-centered: the operator is at the center of the production process and their health, well-being and capabilities are priority alongside business objectives; Synergy with technology: operators interact with advanced technologies such as robots and artificial intelligence, rather than being replaced by them; Resilience and health: the focus is shifting to improving the operator's physical and cognitive resilience and continuous monitoring of their condition.

Functional networks and semi-Markov processes [ 30, 31 ] stand out among the most effective approaches to modeling the human-machine interaction required to implement adaptation procedures. These models allow describing the dialogue as a sequence of states and transitions. The works of scientists of the scientific school of Professor AI Gubinsky have made a significant contribution to the development of these models, showing their applicability to the description, analysis and optimization of dialogue processes [ 32 -37 ]. Directly the problems of adaptive control in human-machine systems were solved in the works [ 38 -39 ].

Unfortunately, despite significant achievements, the problem of operational control of dialogue in a rapidly changing environment taking into account risks such as cyberattacks, deterioration of external conditions and changes in the operator's state is not completely solved.

Problem statement: to develop a conceptual framework for managing the dialogue "manmachine" on the basis of a set of models that should take into account the current state of the manmachine system, as well as possible risks of the external environment and at certain points in time to offer a person the optimal alternative aimed at improving the reliability and safety of management.

3. Results

3.1. An agent-based approach to dialogue management

In connection with the described problems, we propose the implementation of agent-based approach for human operator decision support (Fig.2).

The agent is built on a set of models:

 

Descriptions of the current state of all elements of the human-machine system; Description of the current state of the environment (including risks of negative impacts and cyberattacks);    

Descriptions of alternative variants of human-machine interaction; Estimates of human-machine interaction reliability; Estimation of ergonomic indicators of operator's activity; Selection of optimal variants of human-machine interaction under the conditions of the current state of the system and the environment.

The task of building agents to implement the "Operator 5.0" concept is the global challenge of the next decade.

In the following we will consider only some aspects and examples of building such agents.

3.2. Approach to setting and solving the task of dialogue management

To solve the problem of optimizing human-computer interaction, we developed a bank of optimization models , some of which are described in [ 21, 30-39 ].

In most cases , these models are related to the task of maximizing the probability of error-free and timely execution of functions within the constraints of ergonomic and economic indicators.

The developed methods can and should become the basis for building the agent in question. Further, we will limit ourselves to describing the task of multi-criteria optimization.

In some cases, especially in critical situations, the reliability maximization criterion alone may not be sufficient. Other, often conflicting, criteria, such as speed of execution and operator status, must be taken into account to make better decisions.

Objective: Find the optimal sequence of actions that best satisfies all criteria.

The problem is formulated as a vector optimization problem where: F1(X) – total probability of error-free execution of dialogue procedures; F2(X) – mathematical expectation of the time of implementation of the dialogue interaction process; F3(X) – average tension of the operator's activity throughout the session; X – variable describing the method of dialogue interaction; S – set of combinations of dialogue interaction options.

Assessment of indicators F1(X), F2(X), F3(X) is a rather complex scientific and technical task. We propose that such an assessment be carried out using a "functional network" model, which describes the logical and temporal connections between machine operations and human actions. We have developed some models for automating such assessments while solving other problems, for example, in works [ 32-39 ], but they can also be used to implement a bank of agent-manager models. The impact of adverse conditions and operator stress should be taken into account by recalculating the "control points" based on special risk analysis models.

To do this, we use the procedure “Dynamic generation (recalculation) of initial data”.

The peculiarity is that at each step the transition probabilities and transition times between process states are recalculated depending on the current parameters:   

Rc – risk of cyberattack occurrence;

Re – risk of unfavorable changes in the external environment; α – man-operator stress level.

Since criteria may conflict (eg, increasing speed may increase operator workload), a compromise solution must be found that best meets current priorities.

      

Here we consider only one of the simplest methods based on Thomas Saaty's approach, which naturally does not exhaust the whole bank of optimization models that are included in the agent's model bank.

When the system (dialog management agent) detects a risk and it is necessary to offer the operator an optimal response, it uses a procedure of multi-criteria expert analysis of alternatives based on the method of hierarchy analysis [ 40-43 ].

Sequence of agent actions:

B1. Hierarchy construction. On the basis of the available knowledge base or the dialog with the operator-manager, the agent forms a hierarchy of decisions, where, for example, at the top level is the goal ("Optimization of the dialog session"), at the middle level are the criteria, eg, "Safety", "Speed", "Operator load", and at the bottom level are alternative scenarios for continuing the dialog at the current step; B2. Pairwise comparison of criteria and alternatives. The agent's knowledge base or (in urgent cases) direct instructions from the operator -manager) are used; B3. Automatic synthesis of recommendations. Based on the comparison matrix, the system calculates the priorities for each criterion and then the overall priority for each alternative. The alternative that has the highest priority is selected. This is the optimal action that the system suggests to the operator;

B4. Formation of interactive recommendations to the operator (adaptive interface element).

3.3. Examples of agent-based dialog management implementation

3.3.1. An example from the field of unmanned aerial vehicle control

Consider an operator controlling a unmanned aerial vehicle (UAV) to monitor a critical facility. Normal operation mode:

System state: the UAV is flying along a given route; Operator state: α is at a low level (eg, 2.5 on a 10-point scale); Interface: Standard, displays full set of information: video stream, map, telemetry (speed, altitude, battery charge).

Risk Occurrence. Scenario: The monitoring system detects two threats simultaneously:

 

External environment (Re): The system detects increase in wind and radio interference. Risk Re goes up to 0.8; cyberattack (Rc): Detection of unauthorized access attempts to the UAV control system.

Risk Rc rises to 0.7.

The model, receiving this data, recalculates all parameters. In an operator, due to stress, α starts to increase (eg, to 7.5).

Adaptive management: decision making. The system, using a multi-criteria model and hierarchy analysis method, proposes three alternatives to adjust the UAV's dialog and behavior:   

Alternative A (Safe Return): Return to base via the shortest safe route; Alternative B (Continuation with Enhanced Protection): Continue the mission, but switch to a protected communications link and to manual control to compensate for weather conditions;

Alternative C (Emergency Landing): Make an emergency landing in the nearest safe area.

An example of the input data for analyzing the UAV control strategies when a threat occurs is shown in Fig. 3. and the solution results are shown in Fig. 4.

Adaptive interface. In this scenario, when the system selects Alternative A (Safe Return) because of the high priority, the operator interface adapts instantly to minimize cognitive load and ensure that this action is performed quickly and safely.

The interface changes to reflect the critical situation. The background turns dark red. The video stream is minimized and a large warning appears in the center to return immediately. The auxiliary controls are deactivated, leaving only two active buttons: 

"Confirm Return"; 

"Reject".

This significantly reduces the cognitive load on the operator. The developed agent-based approach for dialog management can be applied in a variety of fields of activity. Despite the fact that e-learning is not directly related to the risks of accidents and disasters, the learning environment can also be aggressive and often carries risks to human health [ 22,44 ].

The formulation of the optimization problem underlying the agent is somewhat different from the above (different objects have their own specifics) and has the following form: where: ɡ – the number for a specific electronic learning module. Each module is different in how it shares information, ; G – how many learning modules are available for the same subject; K(ɡ) – a rating of how comfortable a student feels mentally when using an electronic module ɡ; Xɡ – a variable that describes the type of dialogue used in electronic module ɡ; T(Xɡ) – the average time of implementation of interaction Xɡ; T0 – the longest time you are allowed to train; Z(Xm) – the points a student gets for finishing modulem; z0 – the lowest score you must get on a knowledge test; W(Xɡ) – how difficult the learning process is, in points; w0 – the highest level of difficulty that is allowed; R(ɡ) – all the possible ways to have a dialogue in module a set of combinations of dialogic interaction options for module ɡ.

Fig. 5 shows the principal structure of a manager agent for learning management [ 44 ].

The presented agent uses a neural network to analyze the functional state of a human operator by analyzing the current keyboard handwriting.

4. Testing

The approach has been used many times to build adaptive control systems for complex objects:       control systems for flexible production; chemical enterprise; gas transportation system; banking; e-learning; etc.

5. Conclusions

Modern automated systems are becoming increasingly susceptible to human operator error. People often work under conditions of stress caused by negative environmental influences, cyberattacks and other risks. The "Operator-5.0" concept envisions an increased focus on ensuring human operator comfort, especially in the face of negative influences. To ensure a comfortable humanmachine dialog it is advisable to use an agent-based approach. The agent forming the current information model of the operator and adaptive interface should be based on a set of models such as models for: description of the current state of all elements of the man-machine system; description of the current state of the environment (including the risks of negative impacts and cyberattacks); description of alternative options of man-machine interaction; assessment of the reliability of man-machine interaction; assessment of ergonomic indicators of the operator's activity; selection of optimal variants of man-machine interaction.

Optimization models should take into account as parameters the risks of cyberattacks and deterioration of operating conditions. The arsenal of models should include both single-criteria and multi-criteria problems and provide for automatic changes in interface parameters (depending on current conditions).

The novelty of the results lies in the fact that, in contrast to the known static models of humanmachine interaction optimization, procedures for dynamic adjustment of the interface are proposed, taking into account the current state of the operator, the system and various risks.

Acknowledgements

To the memory of our teachers who founded the scientific school of modeling and optimization of ergotechnical systems, Prof. Anatoly Gubisky, Prof. Vladimir Evgrafov and Prof. Akiva Asherov, we dedicate this article.

The research was supported by the National Research Foundation of Ukraine (Grant Agreement No. 2023.03/0131).

Declaration on Generative AI

The authors have not employed any Generative AI tools.

[10] N. Lathifah and H. -I. Lin, A Brief Review on Behavior Recognition Based on Key Points of Human Skeleton and Eye Gaze To Prevent Human Error, in 13th Asian Control Conference (ASCC), Jeju, Korea,2022, pp. 1396-1403, DOI: 10.23919/ASCC56756.2022.9828066. [11] V. Smiianov et al., Development of informational-communicative system, created to improve medical help for family medicine doctors, Wiadomosci lekarskie ,vol. 71,2 pt 2 ,2018, pp. 331334. [12] Y. Wang et al, Human-machine trust and calibration based on human-in-the-loop experiment, in 2022 4th Int. Conf. SRSE, pp. 476-481, DOI: 10.1109/SRSE56746.2022.10067635. [13] D. Romero and J. Stahre, Towards The Resilient Operator 5.0: The Future of Work in Smart Resilient Manufacturing Systems, Proc. CIRP,vol. 104, 2021,pp. 1089-1094, DOI: https://doi.org/10.1016/j.procir.2021.11.183. [14] D.Mourtzis et al., Operator 5.0: A Survey on Enabling Technologies and a Framework for Digital Manufacturing Based on Extended Reality, J. of Machine Engineering, 22(1), pp.43-69, DOI: https://doi.org/10.36897/jme/147160O. [15] A. Pumpurs, Analysis of Psychographic and Human Factors for Building Personalyzed Product Ecosystems, in: 2021 62nd Int. Sc. Conf. ITMS, 2021, pp. 1-6, DOI:https://doi.org/ 10.1109/ITMS52826.2021.9615277.

References

Ö. Tuttokmaği and A. Kaygusuz , Smart Grids and Industry 4 .0, in Int. Conf. on Art.

Intelligence and Data Processing (IDAP) , 2018 , pp. 1 - 6 , DOI: https://doi.org/10.1109/IDAP. 2018 . 8620887 .

A. Adel , Future of industry 5.0 in society: human-centric solutions, challenges and prospective research areas , J Cloud Comp 11 , 40 ( 2022 ). DOI: https://doi.org/10.1186/s13677- 022-00314-5.

Bereket A. , Hervé P. , Yannick N. , Systemic formalisation of Cyber-Physical-Social System (CPSS): A systematic literature review , Computers in Industry , Vol. 129 , 2021 , 103458, DOI: https://doi.org/10.1016/j.compind. 2021 . 103458 .

F. Sherwani et al., Collaborative Robots and Industrial Revolution 4.0 (IR 4.0) , in 2020 Int.

Conf. ICETST , 2020 , pp. 1 - 5 , DOI: https://doi.org/10.1109/ICETST49965. 2020 . 9080724 .

E.V. Dudukalov et al., Synthesis of the control system for the machinery production using regression analysis in the conditions of industry 4.0, in AIP Conf . Proc. 3021 , 040003 ( 2024 ), DOI: https://doi.org/10.1063/5.0194001.

P. Radanliev et al., Artificial intelligence in cyber physical systems , AI Soc . 2021 ; 36 ( 3 ): 783 - 796 , DOI: 10.1007/s00146-020-01049-0.

E. S. Ogurtsov et al., Microcontroller navigation and motion control system of the underwater robotic complex , ARPN Journal. of Engineering and Applied Sciences , vol. 11 , no.

9, pp. 3110 - 3121 , Jan. 2016 .

A. A. Kolesnikov , A. S. Mushenko and A. D. Zolkin , Two-channel Data Transmission System with Chaotic Generator and Synergetic Observer , in 2019III Int.Conf. CTS ,pp. 46 - 49 , DOI: 10.1109/CTS48763. 2019 . 8973359 .

A. L. Zolkin , I. A. Magomedov and O. V. Kucher , Application of Computer-Aided Technologies for Analysis if Statistical Data of Collectors Wearing Measurements and for Diagnosis of Traction Motors , in 2020 Int. Conf. FarEastCon , 2020 , pp. 1 - 6 , DOI: 10.1109/FarEastCon50210. 2020 . 927107 .

[16] P. I. Paderno , et al., Expert classification: resource-based approach , in: Proc. XXIV Int. Conf. Soft Computing and Measurements (SCM) , 2021 , pp. 31 - 33 . doi: 10 .1109/SCM52931. 2021 . 9507119 .

[17] E. A. Burkov , et al., Analysis of impact of marginal expert assessments on integrated expert assessment , in: Proc. XXIII Int. Conf. Soft Computing and Measurements (SCM) , , 2020 , pp. 14 - 17 . doi: 10 .1109/SCM50615. 2020 . 9198772 .

[18] A. Zgonnikov and G. Markkula , Evidence Accumulation Account of Human Operators' Decisions in Intermittent Control During Inverted Pendulum Balancing , in: 2018 IEEE Int. Conf. SMC , 2018 , pp. 716 - 721 , DOI: https://doi.org/10.1109/SMC. 2018 . 00130 .

[19] O. M. Vultur , Performance Analysis of “Drive Me” - a Human Robot Interaction System , in: 2018 Int. Conf. COMM , 2018 , pp. 125 - 130 , DOI: https://doi.org/10.1109/ICComm. 2018 .8484759

[20] A. P. Rotshtein . Selection of Human Working Conditions Basedon Fuzzy Perfection, J. of Computer and Systems Sciences International , 2018 , Vol. 57 , No. 6 , pp. 927 - 937 .

[21] E. Lavrov , N. Pasko , O. Siryk , Information technology for assessing the operators working environment as an element of the ensuring automated systems ergonomics and reliability , in: Proc. IEEE 15th Int. Conf. Advanced Trends in Radioelectronics , Telecommunications and Computer Engineering (TCSET), Lviv-Slavske, Ukraine , 2020 , pp. 570 - 575 . doi: 10 .1109/TCSET49122. 2020 . 235497 .

[22] O. Burov et al., Cybersecurity in educational networks , IHSI 2020. Advances in Intelligent Systems and Computing , vol. 1131 , pp. 359 - 364 , Springer, Cham , 2020 , DOI: https://doi.org/10.1007/978-3- 030 -39512-4_ 56 .

[23] Y. Zhang et al., Human Response Delay Estimation and Monitoring Using Gamma Distribution Analysis , in 2018 IEEE Int. Conf. SMC , 2018 , pp. 807 - 812 , DOI: https://doi.org/10.1109/SMC. 2018 . 00145 .

[24] L. Fortini et al., A Framework for Real-time and Personalisable Human Ergonomics Monitoring , in: 2020 IEEE/RSJ Int. Conf. IROS , 2020 , pp. 11101 - 11107 , DOI: https://doi.org/10.1109/IROS45743. 2020 . 9341560 .

[25] G. Arvanitis et al., Real time enhancement of operator's ergonomics in physical human - robot collaboration scenarios using a multi-stereo camera system , in: 2023 IEEE International Conference ICIT , Orlando, FL, USA, 2023 , pp. 1 - 6 , DOI: doi: 10.1109/ICIT58465. 2023 . 10143035 .

[26] Y. Watanobe , M. M. Rahman , A. Vazhenin and J. Suzuki , Adaptive User Interface for Smart Programming Exercise , in 2021 IEEE International Conference on Engineering, Technology & Education (TALE) , Wuhan, Hubei Province, China, 2021 , pp. 01 - 07 , doi: 10.1109/TALE52509. 2021 . 9678757 .

[27] D. A. Vidmanov and A. N. Alfimtsev , MARLMUI: Multi-Agent Reinforcement Learning Approach in Mobile Adaptive User Interface , in: 2023 5th International Youth Conference on Radio Electronics , Electrical and Power Engineering (REEPE), , 2023 , pp. 1 - 5 , doi: 10.1109/REEPE57272. 2023 . 10086785 .

[28] C. Haipeng , Y. Li , Y. Jing and Z. Fan, Research on User Experience Adaptive Algorithm in Intelligent Connected Vehicle Interaction , in: 2024 4th International Signal Processing , Communications and Engineering Management Conference (ISPCEM) , Montreal, QC, Canada, 2024 , pp. 340 - 344 , doi: 10.1109/ISPCEM64498. 2024 . 00063 .

[29] I. Yitmen et al., Investigating the Causal Relationships among Enablers of the Construction 5.0 Paradigm: Integration of Operator 5.0 and Society 5 . 0 with Human-Centricity, Sustainability, and Resilience , Sustainability, 2023 , 15 , 9105, DOI: https://doi.org/10.3390/su15119105.

[30] A.N. Adamenko et al., Information-controlling human-machine systems: Research design testing: reference book , M.: Mashinostroenie, pp. 528 , 1993 .

[31] A.I. Gubinsky , A.N. Adamenko , Functional-semantic nets- the universal formalism for defining, designing and estimating the quality of functioning of man-machine systems , in: IFAC Symposia , 1989 , pp. 415 - 420 , DOI: https://doi.org/10.1016/B978-0 -08-036226-7 . 50074 - 0 .

[32] E. Lavrov , N. Pasko , Development of models for computer systems of processing information and control for tasks of ergonomic improvements , in: Information and Software Technologies. ICIST 2018 , Communications in Computer and Information Science 920 ( 2018 ) 98 - 109 . doi: 10 .1007/978-3- 319 -99972- 2 _ 8 .

[33] E. A. Lavrov , et.al., Decision Support Method for Ensuring Ergonomic Quality in Polyergatic IT Resource Management Centers , in 2019 III International Conference on Control in Technical Systems (CTS) , 2019 , pp. 148 - 151 , doi: 10.1109/CTS48763. 2019 . 8973265 .

[34] E. Lavrov , et al., Ergonomics of IT outsourcing. Development of a mathematical model to distribute functions among operators , Eastern-European Journal of Enterprise Technologies 2 ( 2016 ) 32 - 42 . doi: 10 .15587/ 1729 - 4061 . 2016 . 66021 .

[35] E. A. Lavrov , et.al., Decision Support Method for Ensuring Ergonomic Quality in Polyergatic IT Resource Management Centers , in: 2019 III International Conference on Control in Technical Systems (CTS) , 2019 , pp. 148 - 151 , doi: 10.1109/CTS48763. 2019 . 8973265 .

[36] E. Lavrov , N. Pasko , Development of models for computer systems of processing information and control for tasks of ergonomic improvements , in: Information and Software Technologies. ICIST 2018 , Communications in Computer and Information Science 920 ( 2018 ) 98 - 109 . doi: 10 .1007/978-3- 319 -99972- 2 _ 8 .

[37] E.A. Lavrov and O.E. Siryk , "Functional networks for ergonomics and reliability tasks on the 90th anniversary of A. Gubinsky and V. Evgrafov" , Sigurnost , vol. 64 , no. 3 , pp. 245 - 257 , 2022 , DOI: https://doi.org/10.31306/s.64. 3 . 3

[38] E. Lavrov , O. Siryk , Y. Chybiriak , L. Danilova , V. Nahornyi , S. Vakal , A model for the organization of adaptive dialogue interaction 'Man-Computer' taking into account the requirements of reliability and efficiency , in: Proc. IEEE 4th Int. Conf. Advanced Information and Communication Technologies (AICT) , Lviv, Ukraine, 2021 , pp. 31 - 35 . doi: 10 .1109/AICT52120. 2021 . 9628939 .

[39] E. Lavrov , O. Lavrova , Intelligent adaptation method for human-machine interaction in modular E-learning systems , in: CEUR Workshop Proceedings , 2019 , 2393 , pp. 1000 - 1010 .

[40] T.L. Saaty , Decision making for leaders , in IEEE Transactions on Systems, Man, and Cybernetics , vol. SMC- 15 , no. 3 , pp. 450 - 452 , May-June 1985, doi: 10.1109/TSMC. 1985 . 6313384 .

[41] T. L. Saaty , Axiomatic Foundation of the Analytic Hierarchy Process . Management Science , 1986 , 32 ( 7 ), pp. 841 - 855 , https://doi.org/10.1287/mnsc.32.7.841.

[42] E. Lavrov , et al., Human-centered management in polyergatic information systems. Multicriteria distribution of functions between operators , IOP Conf. Ser.: Earth Environ. Sci . 1049 ( 2022 ) 012020 .

[43] I. Svoboda , D. Lande, AI agents in multi criteria decision analysis: automating the analytic hierarchy process with large language models , SSRN , 23 Dec . 2024 . URL: https://ssrn.com/abstract=5069656.

[44] Lavrov , E.A. , Logvinenko , V.G. , Osadchyi , V.V. , Siryk , O.Ye. , Chybiriak , Y.I. : Adaptive learning system based on cognitive independence . In: Advancing lifelong learning and professional development through ICT: insights from the 3L-Person 2023 workshop . CEUR Workshop Proceedings, 2023 , Vol. 3535 , pp. 113 - 127 .