Proceedings of the Institute of Acoustics

 

An underwater passive acoustic serious game to research empathetic artificial intelligence

 

Madalin Facino, University of Bath, UK
Andrei Vladescu, University of Bath, UK
Christof Lutteroth, University of Bath, UK
Alan Hunter, University of Bath, UK
Aaron Roberts, Thales Underwater Systems, UK
Samantha Dugelay, Thales Underwater Systems, UK

 

1 INTRODUCTION AND BACKRGOUND

 

As submarine sensors grow increasingly advanced, managing escalating data volumes and pressure for command teams will be crucial. This surge in data is likely to heighten operator workloads as it necessitates thorough analysis and integration into the tactical overview⁶⁰. Workload, in this sense, can be understood as the balance between the mental processing demanded by a task and the operator’s capability to manage this demand^{18 19}. Maintaining an optimal cognitive arousal level is vital⁸; overstimulation, or cognitive overload, can result in risk-taking behaviors³⁸, disordered task execution¹⁹, delayed task completion⁵, overlooked secondary tasks⁴⁹, vigilance decrement³⁶, diminished performance³⁴, and, in severe cases, complete performance failure¹⁹. Historical incidents, like the 2015 collision between the stern trawler Karen (B317) and a Royal Navy submarine, underline the grave consequences of operator overload³⁰. 

 

Sonar operations in underwater settings face unique challenges due to complex marine conditions, weak signal reception compounded by strong background noise, and significant interference—all of which degrade communication quality and complicate underwater object detection^{47 62 63}. These conditions are exacerbated by high transmission delays, limited bandwidth, and the energy constraints of underwater sensor nodes⁵⁸. Further, the psychological toll on sonar operators is profound. The complexity of natural sonar targets generates multiple echo points, increasing the cognitive demands of distinguishing between actual targets and clutter^{39 52 61}. Long-term exposure to such intricate auditory signals can affect operators’ psychological well-being, evident in changes to auditory event-related potentials³². These factors, combined with the inherent stress of submarine environments⁵⁰, intensify the psychological strain on operators⁵⁵. 

 

Whilst the collaboration between sonar operators and artificial intelligence (AI) is becoming increasingly necessary with data growth, distrust in AI technologies remains a major obstacle^{22 26 37}. This paper proposes a method for testing a novel strategy to alleviate these psychological impacts and trust barriers through the development of empathetic AI. We suggest that integrating empathy into AI could foster trust, thereby enhancing the team dynamic, team performance, and reducing operator cognitive load. 

 

To explore this hypothesis, we have developed a serious game within the Unity platform, simulating realistic sonar operation conditions. This game focuses on key psychological and cognitive factors involved in sonar operations whilst utilizing established acoustic models and physics^{1 6 59}. The game design is centered around tasks that involve perception, multitasking, and stress management, ensuring that our study maintains ecological validity⁵⁷. 

 

Submarine sonar operations pose a triad of formidable challenges for operators: multitasking, perception, and pressure management, each significantly affecting operational efficiency and psychological well-being^{25 57}. The intense demands of multitasking require operators to simultaneously process varied acoustic signals, maintain situational awareness, communicate findings, and make pivotal decisions. This cognitive burden, as underscored by findings using NASA’s Task Load Index^{18 19}, highlights the pressing need for innovative tools designed to alleviate these pressures. 

 

The perceptual challenges in sonar operations are equally daunting. Operators must distinguish critical signals from pervasive background noise within a complex underwater soundscape⁵³. This task demands not only technical acumen but also continuous cognitive engagement. Additionally, the high pressure conditions inherent in submarine operations—characterized by extended submersion, erratic sleep patterns, and high-stakes decision-making – intensify psychological stress, potentially undermining both performance and mental health⁵⁰. 

 

Current training simulations like MATB⁴⁵ and the naval warfare simulation game Dangerous Waters⁴³ offer foundational frameworks but do not fully address these multifaceted challenges. MATB focuses primarily on procedural and multitasking skills but offers limited insight into the perceptual complexities of sonar operations. Dangerous Waters, while providing an intricate portrayal of naval engagements, is constrained by its complexity and steep learning curve, which may impede its effectiveness in psychological impact studies focused on specific cognitive and emotional responses. Critically, these platforms do not incorporate biosignal monitoring, a key aspect in assessing physiological markers during operations^{16 17 21} which will serve to inform an empathetic response. 

 

Recognizing the state-of-the-art in current underwater acoustic training tools, there is a critical need for a simulation that not only represents the acoustic and operational conditions of sonar operations but also serves as a dedicated medium for studying psychological impacts. Our proposed serious game environment, which will integrate empathetic AI, is designed to address this need. By recognizing and responding to human emotions, empathetic AI holds the promise of enhancing human-machine collaboration. This innovative integration could reduce cognitive load, improve perceptual accuracy through adaptive interfaces, and alleviate operational pressures by offering psychologically attuned support. Furthermore, the efficacy of serious game environments in such applications has been substantiated in an extended abstract we published in the ACM Library through the First International Symposium on Trustworthy Autonomous Systems (TAS23)¹³. 

 

This interdisciplinary effort combines psychology, computer science, and engineering to not only test our hypotheses about empathetic AI but also to facilitate the development of more intuitive and empathetic user and AI interfaces. By simulating representative real-world conditions within a controlled setting, we aim to provide insights that could lead to significant advancements in human-machine collaboration. 

 

 

2 DESIGN AND DEVELOPMENT OVERVIEW

 

Our serious game, based in Unity due to its robust framework and relative ease of use, is currently in its prototyping and refining stage. It simulates the critical aspects of submarine sonar operations for psychological studies — balancing acoustic realism with an examination of psychological impacts. Central to the game’s narrative is a mission to intercept submarines, depicted as ‘poachers’ using active sonar, which poses a threat to whale populations³⁵ . This scenario frames the gameplay within realistic operational tasks while avoiding sensitive real-world conflicts.

 

2.1 User Interface and Aesthetic Choices

 

The user interface harmonizes the roles of sonar operator (SOP), officer of the watch (OOW), and target motion analyst (TMA)^{18 49} , presenting operators with an array of interactive tools. Operators navigate through a time-bearing chart for real-time tracking of acoustic signals (left side of Figure 1), employ a top-down submarine view to direct the sonar beam (center of Figure 1), and utilize a spectrogram (right side of Figure 1) for intricate sound frequency analysis. Additionally, operators will wear headphones during the study and will be able to listen to the audio being displayed on the screen and hear it being impacted by their inputs in real-time. The game offers operators the option to make the weighted decision of apprehending the target (upper center of Figure 1), pushing them to fully ensure their selected entity is in fact a target. Tactical decision-making is enhanced through the use of trackers to label potential targets (upper left of Figure 1). The interface is complemented by an oracle AI system that proposes classifications and tracks for the acoustic signals, which players can adjust, fostering engagement and gradually building trust in the technology.

 

 

Figure 1: Game Interface

 

Aesthetically, the game uses gamified symbology and coloring to facilitate accessibility for university participants. Future iterations will incorporate NATO standard colors and symbols³ and will utilize fonts optimized for readability⁶³. Color schemes are carefully chosen to induce stress during critical gameplay moments^{20 29 44}, such as red flashing lights as the timer runs down, maintaining a balance between challenge, gamification, and realism.

 

2.2 AI Design and Evolution 

 

The AI in the game is an Oracle (or all knowing) in order to maintain total controllability and replicability. We degrade the performance of the Oracle intentionally to simulate realistic AI behavior. It serves as an informed assistant, having a comprehensive overview of the game state and aiding in sound identification and classification. Initially, the AI marks potential tracks on the time-bearing chart with varying degrees of pre-programmed confidence¹² in relation to the given scenario, encouraging players to interact with and trust¹² the AI’s judgment through transparency⁵⁶. This feature is designed to be user-overridable, allowing for dynamic interaction and gradual trust-building in the AI’s capabilities. Feedback from initial pilot sessions will guide the AI’s development, particularly toward integrating more sophisticated empathetic responses. Future enhancements will include the necessary integration of an empathetic feedback loop. Our AI currently works without a feedback loop for piloting purposes but a crucial step in developing our environment for impact studies will be to integrate a biosignal feedback loop — which has been designed and validated by colleagues at the University of Bath and which will be published at CHI2024¹¹. With such integration, we will be able to have our AI change its form of support depending on the operator’s cognitive state. Additionally, we plan to integrate more advanced intentional AI performance variability, such as occasional failures and uncertainties in distinguishing between whale calls and other sounds⁴⁸ or more regular changing of confidence, to enrich the training scenario, better understand the impact of performance on trust, and better mimic real-world conditions. 

 

2.3 Ecological Validity and Scenario Design 

 

To ensure ecological validity while focusing on the psychological impacts, the simulation strategically omits extraneous technical details that do not directly contribute to our core research objectives. This simplification allows us to concentrate on critical cognitive domains relevant to sonar operations —perception, multitasking, and stress management⁵⁷ — without overwhelming operators with unnecessary information. 

 

The game is designed to simulate varying degrees of operational demand and stress¹⁸. Scenarios increase in complexity and environmental pressure, challenging the operator’s cognitive load and decision making abilities under different stress levels. Stress-inducing events are carefully timed to test the operator’s ability to maintain situational awareness and make precise judgments, reflecting the pressures of real sonar operations. A scoring system mimics the high stakes of operational decisions, where errors can lead to significant consequences⁴². 

 

Through these structured design elements, the simulation serves both as a technological and psychological exploration tool, and as an immersive representation of the challenges faced by sonar operators. This deliberate design ensures that the environment is not only a test bed for advanced AI interactions but also a platform for operational and psychological research. 

 

3 TECHNICAL IMPLEMENTATION 

 

3.1 Pre-Processing 

 

Prior to the start of the simulation, pre-computation is undertaken to ensure accurate acoustic modeling and integration into the Unity environment. Each entity — whether it be a ship, target, or marine life — is assigned a specific audio signal (.wav file) selected from an underwater acoustic database^{2 10 24 46}. These entities are given defined paths relative to a stationary submarine, as depicted in Figure 3, which are plotted to ensure realistic spatial navigation. The audio for these entities is adjusted using a validated Bellhop propagation model^{40 54}, which applies the Munk sound speed profile¹ to realistically reflect the sound propagation across various environmental conditions such as distance, depth, and sound speed. The effects of the propagation model can be seen in Figure 2(c) (in relation to Figure 3) as the sound of the whale intensifies in the first few minutes as it approaches the submarine and then dissipates as the whale moves further away. 

 

 

Figure 2: Propagation Model 54 Example.

 

 

Following the path definition for each entity, we establish the AI’s behavior in advance, programming it to react with varying degrees of confidence to different audio characteristics as they unfold within the scenario. By pre-setting the AI’s confidence levels¹² at timed intervals, we simulate a dynamic and realistic response to complex acoustic environments. For instance, the AI’s confidence in identifying a source might be programmed to decrease when that source crosses paths with another entity, reflecting the increased difficulty in distinguishing between overlapping acoustic signals²⁸ (see Figure 3). This aspect enhances the simulation’s realism, providing players with a more nuanced and challenging interactive experience.

 

 

Figure 3: Path Planning Process for First Pilot Run

 

In addition to source handling, ambient noise modeling is also a crucial part of the pre-computation. Noise power spectral density⁹, based on the Wenz curves⁵⁹ for a given sea state, is used to generate a consistent audio signal for the ambient noise. This ensures the noise complements the dynamic acoustic environment of the simulation, providing a realistic background of sea-state-specific sound. 

 

3.2 Real-Time Computation 

 

Once the pre-computation is complete, the simulation transitions to real-time execution within Unity, where these elements are dynamically integrated to create an immersive user experience. Each precomputed audio source and noise signal is introduced into Unity as an ”AudioEntity” object. The relative positions of these sources are translated into Unity units, mapping the realistic 3 x 3 km world region onto a 100 x 100 unit grid within the game environment. Unity’s native sound propagation model is disabled in favor of our precomputed data, which is tailored to offer a more sophisticated and accurate acoustic representation. 

 

During gameplay, players can adjust settings such as the sonar beam’s direction, width²⁷, and bandwidth⁵¹. These user inputs dynamically influence the audio processing of both the source signals and ambient noise, reflecting the changes in the acoustic environment in real time. The modification of the sonar settings adjusts the visibility of sources and the level of noise based on the beam’s width and the selected frequency bandwidth. 

 

This dynamic scaling is also applied to the ambient noise AudioEntity (which is anchored to the center of the beam to maintain consistent noise through Unity’s audio listener), ensuring that the noise levels and source visibility are appropriately adjusted according to the player’s settings. All of our AudioEntities are then put into the environment and are heard and displayed on the spectrogram, through Unity’s spectrogram fucntionality, when within the user-selected beamwidth and bandwidth. 

 

The real challenge arises in handling the time-bearing plot, which is not natively supported by Unity. For each time interval, signals extracted from the real-time spectrogram are allocated to appropriate bearing bins to visually represent the directionality and intensity of underwater sounds. Gaussian noise²³, generated based on the variance calculated from the adjusted noise level, is artificially introduced to simulate realistic ambient noise in the time-bearing plot. 

 

The integration of precomputed acoustic data and the real-time dynamic adjustments in Unity form the backbone of our simulation. This sophisticated approach ensures a high fidelity simulation that is both immersive and scientifically based, enhancing the realism and educational value of the acoustic experience. 

 

4 HUMAN FACTORS AND COGNITIVE DOMAINS 

 

4.1 Studying Operator Engagement and Trust in Automated Systems 

 

Our simulation aims to provide a unique perspective on human-machine interaction, particularly in the specialized domain of underwater acoustics. It offers a platform for studies into how operators engage with sonar interfaces and AI systems, focusing on key psychological aspects such as trust, reliance, and the effectiveness of AI in supporting complex decision-making processes and reducing cognitive overload^{8 41}. By integrating biosignal and eye-tracking technologies⁷, alongside thorough post-study analyses, we will be able to measure when and how operators respond to AI recommendations. This data will then be used for developing optimal strategies for human-machine collaboration, and for understanding the broader implications on user fatigue, performance, stress, and cognitive load^{16 17 21 33}. 

 

4.2 Analyzing the Impact of Empathetic AI 

 

At the heart of our research lies a deep analysis of the psychological stressors inherent in submarine sonar operations. The serious game simulates these high-pressure environments, providing a controlled laboratory for studying the impacts of multitasking, perceptual challenges, and operational stress on psychological well-being. By varying environmental complexity and levels of AI support, we plan to methodically assess their effects on cognitive load¹⁴, stress, and operator performance. In the future we may additionally work to further replicate such stressful conditions by creating more immersive sensory environments outside of the simulation^{17 31}. 

 

Empathetic AI stands as a cornerstone of our experimental framework, aimed not only to enhance contact identification and classification but also to adjust its responses based on real-time operator monitoring. Our research design will critically examine three distinct conditions — no AI, neutral AI, and empathetic AI — to ascertain the potential of empathetic AI in enhancing human-machine collaboration. This design will be repeated with different expressions of empathy to build upon our understanding. Our aim is to provide support that is both perceptively and psychologically attuned to the operators’ needs, thus fostering more effective and intuitive human-machine interactions — building trust¹⁵ and a more optimal human-machine team dynamic. 

 

Upon the completion of initial piloting phases, our first primary study will implement a within-subjects design, engaging a diverse cohort of at least 60 participants to ensure a broad analysis across various conditions, thereby maximizing ecological validity. The study will be a 3 x 2 design using the above experimental conditions — at both low and high demand to evaluate the effects of empathetic AI integration on operational performance. We will adopt a continuous gameplay approach which moves beyond tra ditional win/lose frameworks by allowing participants to accumulate points that influence their financial compensation⁴², thereby adding a realistic element of performance-related stress. A preliminary phase will familiarize participants with the game mechanics before formal data collection commences to ensure there is no practice effect⁴ acting as a confounding variable, supported by a detailed screening questionnaire to control for factors such as experience level and personality traits — ensuring methodological rigor. 

 

5 CONCLUSION 

 

In conclusion, the integration of an acoustic model into the Unity environment has been successfully achieved, with an interface that carefully aims to balance realism, gamification, and ecological validity. The implemented scenarios and interface are crafted to maintain this equilibrium, providing a foundation for ongoing refinements through initial testing phases. This work not only sets the stage for future investigations into the efficacy of empathetic AI but also employs an oracle AI as a proxy, simulating behaviors that pave the way for the development of future empathetic AI systems. These efforts collectively work to ensure a comprehensive exploration of how psychologically-attuned AI can enhance human-machine interactions in complex decision-making environments. 

 

REFERENCES 

 

1. A Deep Water Problem: the Munk Profile.

2. Historic Naval Sound and Video.

3. APP-6(C) NATO Joint Military Symbology. NATO Unclassified, May 2011.

4. Ralph H. B. Benedict and Dennis J. Zgaljardic. Practice Effects During Repeated Administrations of Memory Tests With and Without Alternate Forms. Journal of Clinical and Experimental Neuropsychology, 20(3):339–352, June 1998. Routledge. https://doi.org/10.1076/jcen.20.3.339.822

5. Francesco N. Biondi, Angela Cacanindin, Caitlyn Douglas, and Joel Cort. Overloaded and at Work: Investigating the Effect of Cognitive Workload on Assembly Task Performance. Human Factors, 63(5):813–820, August 2021. SAGE Publications Inc.

6. Michel Bouvet and Stuart C. Schwartz. Underwater noises: Statistical modeling, detection, and normalization. The Journal of the Acoustical Society of America, 83(3):1023–1033, March 1988.

7. D. Bruneau, M. A. Sasse, and J. D. McCarthy. The Eyes Never Lie: The Use of Eyetracking Data in HCI Research. In Proceedings of CHI 2002: Conference on Human Factors in Computing Systems, Minneapolis, Minnesota, USA. ACM, 2002.

8. John Buchanan and Ned Kock. Information Overload: A Decision Making Perspective. In Murat Koksalan and Stanley Zionts (eds.), Multiple Criteria Decision Making in the New Millennium, pp. 49–58. Springer, Berlin, Heidelberg, 2001.

9. Carlos Cordeiro, Dave Cavalcanti, and Saishankar Nandagopalan. Cognitive radio for broadband wireless access in TV bands: The IEEE 802.22 standards. In Alexander M. Wyglinski, Maziar Nekovee, and Y. Thomas Hou (eds.), Cognitive Radio Communications and Networks, pp. 387–429. Academic Press, Oxford, January 2010.

10. Lucas Domingos. VTUAD: Vessel Type Underwater Acoustic Data., September 2022.

11. Dominic Potts, Zoe Broad, Tarini Sehgal, Joseph Hartley, Eamonn O’Neill, Crescent Jicol, Christopher Clarke, and Christof Lutteroth. Sweating the Details: Emotion Recognition and the Influence of Physical Exertion in Virtual Reality Exergaming., March 2024.

12. Daniel C. Elton. Self-explaining AI as an Alternative to Interpretable AI. In Ben Goertzel, Aleksandr I. Panov, Alexey Potapov, and Roman Yampolskiy (eds.), Artificial General Intelligence, pp. 95–106. Springer International Publishing, Cham, 2020.

13. Madalin Facino, Alan Hunter, Christof Lutteroth, Aaron Roberts, and Samantha Dugelay. Storyboarding a Serious Game Environment for Evaluating the Impact of Empathetic AI for Sonar Operators. In Proceedings of the First International Symposium on Trustworthy Autonomous Systems (TAS ’23), pp. 1–2. Association for Computing Machinery, New York, USA, July 2023.

14. Julia R. Fox, Byungho Park, and Annie Lang. When Available Resources Become Negative Resources: The Effects of Cognitive Overload on Memory Sensitivity and Criterion Bias. Communication Research, 34(3):277–296, June 2007. SAGE Publications Inc.

15. Patrick Gebhard, Ruth Aylett, Ryuichiro Higashinaka, Kristiina Jokinen, Hiroki Tanaka, and Koichiro Yoshino. Modeling Trust and Empathy for Socially Interactive Robots. In Juliana Miehle, Wolfgang Minker, Elisabeth André, and Koichiro Yoshino (eds.), Multimodal Agents for Ageing and Multicultural Societies: Communications of NII Shonan Meetings, pp. 21–60. Springer, Singapore, 2021.

16. Giorgos Giannakakis, Dimitris Grigoriadis, Katerina Giannakaki, Olympia Simantiraki, Alexandros Roniotis, and Manolis Tsiknakis. Review on Psychological Stress Detection Using Biosignals. IEEE Transactions on Affective Computing, 13(1):440–460, January 2022.

17. Kunal Gupta, Ryo Hajika, Yun Suen Pai, Andreas Duenser, Martin Lochner, and Mark Billinghurst. In AI We Trust: Investigating the Relationship between Biosignals, Trust and Cognitive Load in VR. In Proceedings of the 25th ACM Symposium on Virtual Reality Software and Technology, VRST ’19, pages 1–10, New York, NY, USA, November 2019. Association for Computing Machinery.

18. Sophie M Hallam, Neville A Stanton, Aaron P J Roberts, Kiome A Pope, and Daniel Fay. The effects of a circle room configuration on submarine command team workload. Human Factors, 2022.

19. Sandra G. Hart and Lowell E. Staveland. Development of NASA-TLX (Task Load Index): Results of Empirical and Theoretical Research. In Peter A. Hancock and Najmedin Meshkati, editors, Advances in Psychology, volume 52 of Human Mental Workload, pages 139–183. North-Holland, January 1988.

20. Takeshi Hatta, Hirotaka Yoshida, Ayako Kawakami, and Masahiko Okamoto. Color of Computer Display Frame in Work Performance, Mood, and Physiological Response. Perceptual and Motor Skills, 94(1):39–46, February 2002. Publisher: SAGE Publications Inc.

21. Seyyed Abed Hosseini, Mohammad Ali Khalilzadeh, and Sahar Changiz. Emotional stress recognition system for affective computing based on bio-signals. Journal of Biological Systems, 18(spec01):101–114, October 2010. Publisher: World Scientific Publishing Co.

22. Wolfgang Kersten, Carlos Jahn, Thorsten Blecker, and Christian M Ringle. Prevailing technologies and adoption obstacles in maritime logistics.

23. Bruce Kessler. Gaussian White Noise.

24. Chris Knowlton. North Atlantic Right Whale.

25. David Kobus and Lawrence Lewandowski. Critical Factors in Sonar Operation: A Survey of Experienced Operators. DTIC, December 1991.

26. Rafal Kocielnik, Saleema Amershi, and Paul Bennett. Will you accept an imperfect AI?: Exploring designs for adjusting end-user expectations of AI systems. pages 1–14, April 2019.

27. Nikolai Kolev. Sonar Systems. BoD – Books on Demand, September 2011. Google-Books-ID: DeDkxwEACAAJ.

28. Manav Kumar and Sharifuddin Mondal. Recent developments on target tracking problems: A review. Ocean Engineering, 236:109558, September 2021.

29. Teresa Kutchma. The Effects of Room Color on Stress Perception: Red versus Green Environments. Journal of Undergraduate Research at Minnesota State University, Mankato, 3(1), August 2014.

30. Marine Accident Investigation Branch. Collision between the stern trawler Karen and a dived Royal Navy submarine, October 2016.

31. Miguel Melo, Hugo Coelho, Guilherme Gonçalves, Nieves Losada, Filipa Jorge, Mário Sérgio Teixeira, and Maximino Bessa. Immersive multisensory virtual reality technologies for virtual tourism. Multimedia Systems, 28(3):1027–1037, June 2022.

32. L. Merrill and D. Kobus. The effect of sonar experience and age on the auditory event-related potential. 1993.

33. Nick Merrill and Coye Cheshire. Trust Your Heart: Assessing Cooperation and Trust with Biosignals in Computer-Mediated Interactions. In Proceedings of the 2017 ACM Conference on Computer Supported Cooperative Work and Social Computing, CSCW ’17, pages 2–12, New York, NY, USA, February 2017. Association for Computing Machinery.

34. Justin Owens, Thomas Dingus, Feng Guo, Youjia Fang, Miguel Perez, and Julie McClafferty. Crash Risk of Cell Phone Use While Driving: A Case-Crossover Analysis of Naturalistic Driving Data. Technical report, Virginia Tech Transportation Institute, January 2018.

35. E. C. M. Parsons. Impacts of Navy Sonar on Whales and Dolphins: Now beyond a Smoking Gun? Frontiers in Marine Science, 4, September 2017. Publisher: Frontiers.

36. Nathalie Pattyn, Xavier Neyt, David Henderickx, and Eric Soetens. Psychophysiological investigation of vigilance decrement: Boredom or cognitive fatigue? Physiology & Behavior, 93(1):369–378, January 2008.

37. Andrew Pfau. A Roadmap to Successful Sonar AI. Center for International Maritime Security, August 2021.

38. Gloria Phillips-Wren and Monica Adya. Decision making under stress: the role of information overload, time pressure, complexity, and uncertainty. Journal of Deci. sion Systems, 29(sup1):213–225, August 2020. Publisher: Taylor & Francis eprint: https://doi.org/10.1080/12460125.2020.1768680.

39. J. Pitton, S. Philips, L. Atlas, J. Ballas, D. Brock, B. McClimens, and M. Miller. Aural classification of impulsive-source active sonar echoes. The Journal of the Acoustical Society of America, 119:3394–3395, 2006.

40. Michael B. Porter. The BELLHOP Manual and User’s Guide.

41. Iain Rice and David Lowe. A Decision Support System to Ease Operator Overload in Multibeam Passive Sonar. IEEE Journal of Oceanic Engineering, 44(1):100–110, January 2019.

42. Ganit Richter, Daphne R. Raban, and Sheizaf Rafaeli. Studying Gamification: The Effect of Rewards and Incentives on Motivation. In Torsten Reiners and Lincoln C. Wood, editors, Gamification in Education and Business, pages 21–46. Springer International Publishing, Cham, 2015.

43. Aaron Roberts, Neville Stanton, and Daniel Fay. The Command Team Experimental Test-bed Stage 1: Design and Build of a Submarine Command Room Simulator. Procedia Manufacturing, 3:2800–2807, January 2015.

44. Julie Saint-Lot, Jean-Paul Imbert, and Frédéric Dehais. Red Alert: A Cognitive Countermeasure to Mitigate Attentional Tunneling. In Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems, CHI ’20, pages 1–6, New York, NY, USA, April 2020. Association for Computing Machinery.

45. Yamira Santiago-Espada, Robert R. Myer, Kara A. Latorella, and James R. Comstock. The Multi Attribute Task Battery II (MATB-II) Software for Human Performance and Workload Research: A User’s Guide. Technical Report L-20031, July 2011. NASA Langley Research Center, Stinger Ghaffarian Technologies, Inc. (SGT, Inc.). NTRS Document ID: 20110014456.

46. David Santos-Domínguez, Soledad Torres-Guijarro, Antonio Cardenal-Lopez, and Antonio Peña-Gimenez. ShipsEar: An underwater vessel noise database. Applied Acoustics, 113:64–69, December 2016.

47. Pengfei Shi, Xinnan Fan, Jianjun Ni, Zubair Khan, and Min Li. A novel underwater dam crack detection and classification approach based on sonar images. PLOS ONE, 12(6):e0179627, June 2017. Publisher: Public Library of Science.

48. Won Shin, Da-Sol Kim, and Hyunsuk Ko. Target Tracking from Weak Acoustic Signals in an Underwater Environment Using a Deep Segmentation Network. Journal of Marine Science and Engineering, 11(8):1584, August 2023.

49. Neville A. Stanton, Aaron P. J. Roberts, and Daniel T. Fay. Up periscope: understanding submarine command and control teamwork during a simulated return to periscope depth. Cognition, Technology & Work, 19(2):399–417, September 2017.

50. Peter Suedfeld and G. Daniel Steel. The Environmental Psychology of Capsule Habitats. Annual Review of Psychology, 51:227–253, February 2000.

51. Dajun Sun, Kaiyang Hou, and Tingting Teng. Underwater weak moving target detection method based on wideband Multi-pulse coherent integration. Applied Acoustics, 212:109616, September 2023.

52. J. Sutlive and R. Muller. Dynamic echo signatures created by a biomimetic sonar head. Bioinspiration & Biomimetics, 14:066014, 2019.

53. A. Testolin and R. Diamant. Combining denoising autoencoders and dynamic programming for acoustic detection and tracking of underwater moving targets. Sensors, 20:2945, 2020.

54. Alfie Treloar. Passive Acoustic Array Modelling and Processing for a Wave-Propelled Unmanned Surface Vessel. PhD thesis, University of Bath, April 2022.

55. Charles Van Wijk and Vittorio Dalla Cia. “Covert Coping” in Extreme Environments: Insights from South African Submarines. Journal of Human Performance in Extreme Environments, 12(2), July 2016.

56. Warren J. von Eschenbach. Transparency and the Black Box Problem: Why We Do Not Trust AI. Philosophy & Technology, 34(4):1607–1622, December 2021.

57. Wenbi Wang. Development of a Baseline Workload Model for Future Submarine Sonar Crewing Analysis. Proceedings of the Human Factors and Ergonomics Society Annual Meeting, 60(1):670–674, September 2016. Publisher: SAGE Publications Inc.

58. Xinxin Wang, Danyang Qin, Min Zhao, Ruolin Guo, and Teklu Merhawit Berhane. UWSNs positioning technology based on iterative optimization and data position correction. EURASIP Journal on Wireless Communications and Networking, 2020(1):158, August 2020.

59. Gordon M. Wenz. Acoustic ambient noise in the ocean: Spectra and sources. The Journal of the Acoustical Society of America, 34(12):1936–1956, 1962.

60. D. D. Woods, E. S. Patterson, and E. M. Roth. Can We Ever Escape from Data Overload? A Cognitive Systems Diagnosis. Cognition, Technology & Work, 4(1):22–36, April 2002.

61. V. Young and P. Hines. Perception-based automatic classification of impulsive-source active sonar echoes. The Journal of the Acoustical Society of America, 122:1502–1517, 2007.

62. Chengpengfei Zhang, Guoyin Zhang, Heng Li, Hui Liu, Jie Tan, and Xiaojun Xue. Underwater target detection algorithm based on improved YOLOv4 with SemiDSConv and FIoU loss function. Frontiers in Marine Science, 10, March 2023.

63. Hanwen Zhang, Zhen Qin, Yichao Zhang, Dajiang Chen, Ji Gen, and Hao Qin. A practical underwater information sensing system based on intermittent chaos under the background of Lévy noise. EURASIP Journal on Wireless Communications and Networking, 2022(1):41, May 2022.