Skip to article
AHFE AHFE Discoveries AHFE International Open Access Journal
AHFE OPEN ACCESS OPEN ACCESS CONFERENCE PROCEEDINGS PAPER

New Forms of Alliances Between Humans and Technology: The Role of Human Agency to Enhance the Design and Setting Up of Artificial Intelligence Based Learning Tools

Publication year
Digital Object Identifier10.54941/ahfe1003154

Abstract

This paper investigates the development of new forms of alliances between humans and technology in the context of interaction and collaboration between human and artificial intelligence (AI), with the aim to address responsible development and use of artificial intelligence-based learning tools (adaptive learning).The future of adaptive learning can make important contributions in education and improve student and teacher experiences, while raising challenges and risks that can be formulated as impacts on human life, both at the individual and societal levels. In todays’ context, adaptive learning tools are implemented to support instructors in the delivery of content and to provide learners personalised learning, without necessary involving end users throughout the design and running phases of these tools. The findings indicate there is a need to focus on human agency and pedagogy to enhance the adoption of adaptive learning. This paper investigates whether an embedded feedback system in AI based learning tools, based on participatory design, involving learners and instructors would help to overcome the risks and challenges related to the adoption of these tools. A process that involves an embedded feedback system is proposed. The approach adopted is to extend research on Cybernetics theory and other research on human-AI interactions to revise mutualistic relationship between humans and AI. The theory of cybernetics brings the idea to control entropy in a system through feedback. In other words, cybernetics is the science of feedback, information that travels from a system through its environment and back to the system (Wiener 1961). As stated by Dubberly, learning and adapting to the needs of a system are the goals of both iterative design processes and cybernetics. Thus, Cybernetics can provide a useful framework for augmenting designers in creating human-centred interactive AI-enabled products (Dubberly 2015). As these systems become more reliable and easier to work with, designers can embed them into tools-as AI subsystems-that can interact with people. This interaction can then support the AI subsystem to further learn and adapt to the user.This research will further extend with empirical studies to investigate more details on how human computer interaction design approaches can address risks and challenges associated with it and to show that there is a need to focus on human agency to enhance the design and setting up of adaptive learning, where learners, instructors and other education stakeholders need to be actively involved. The methodology used for data collection is primarily interviews conducted with different groups of AI-enabled tools practioners (data scientists, software engineers, software vendors, and leaders within organisations, EdTech professionals). Semi-structured interviews are chosen. The methodology used for data analysis is thematic analysis. This study is further supported with extensive literature reviews on research conducted on cybernetics, education technology and adaptive learning, and analysis of case-studies of organisations in the process of adoption of adaptive learning systems.Dubberly, H., Pangaro, P.: Cybernetics and design: conversations for action. Cybern. Hum. Know. 22(é-3), 73–82 (2015)Wiener, N.: Cybernetics or Control and Communication in the Animal and Machine, vol. 25. MIT Press (1961)

INTRODUCTION

The adoption of AI based learning tools, defined as adaptive learning tools, has evolved with the development of new forms of AI. Adaptive learning refers broadly to a learning process where the content taught or the way such content is presented changes, or “adapts”, based on the responses of the individual learner (Oxman and Wong 2014). Adaptive learning has the poten- tial to improve learners’ and instructors’ experiences and make an important impact on education while raising risks and challenges that can be framed on human life, both in individual and societal level. According to interviews conducted with AI-based tools practitioners, these risks and challenges can be grouped at three level; i) ethical (related to the usage of learning data without the consent of users), ii) educational process (difficulty to integrate AI into the learning processes), iii) implementation at scale (shortage of skills).

In today’s environment, adaptive learning tools are being implemen- ted to support instructors in content delivery and to provide learners

© 2023. Published by AHFE Open Access. All rights reserved. 86

New Forms of Alliances Between Humans and Technology 87

withpersonalised learning, without necessarily involving them throughout the design and execution phases of these tools.

This paper investigates whether the development of new forms of intera- ction and collaboration between AI-based learning tools and its users, based on participatory design, would address risks and challenges linked to the adoption of these tools. The proposed approach is to analyse the cybernetics theory as a potential framework for analysing human-AI interactions and augmenting designers in creating human-centred interactive AI-enabled pro- ducts. Cybernetics is defined as the science of feedback, and appears as part of the earliest study on human-machine interactions (Wiener, 1960).

Participatory design, with a need to focus on human agency in the design and running of AI-based learning tools is presented as primordial, to address risks and challenges related to the adoption of these tools. In the context of human-AI alliances, designers can embed participatory design approaches into tools -as AI subsystems- that can interact with humans (Dubberly 2015). This interaction can then support the AI subsystem to further learn and adapt to the user.

The methodology used for data collection is based on extensive literature review on cybernetics theory, education technology and adaptive learning; supported by interviews conducted with different group of AI based-tools practitioners and analysis of case-studies of organisations in the process of adoption of adaptive learning systems.

NEW FORMS OF ALLIANCES BETWEEN HUMAN AND TECHNOLOGY Adaptive learning tools, based on new forms of AI technologies, can be of various types, ranging from simple systems with preconceived set of rules to complex systems with self-learning algorithms (Brinton et al. 2015). These tools are designed and developed by experts, without necessarily the invo- lvement of learners and instructors. The findings indicate that a focus on human agency throughout the design and running phases of adaptive lear- ning tools where learners, instructors and other education stakeholders are actively involved can have an important impact in addressing the risks and challenges associated with the adoption of AI-based tools (Osmanoglu 2022). Hence, the concept of participatory design for AI-based tools plays an impor- tant role in the development of adaptive learning systems. The analysis of participatory design requires a focus on forms of human-AI interactions and collaborations, that involve embedded feedback systems within adaptive lear- ning tools. The following section presents various forms of alliances between humans and AI that can support participatory design. The approach adopted is to present cybernetics theory as a framework that can be used to support research and development on new forms of human-AI alliances.

CYBERNETICS The notion of cybernetics has been introduced by Norbert Wiener, as as “the scientific study of control and communication in the animal and the

88 Osmanoglu

machine”, marking the early studies on human-machine interactions (Wie- ner 1961). The theory of cybernetics introduces the idea to control entropy in a system through feedback (Wiener 2019). In other words, cybernetics is the science of feedback, a response within a system that influences the conti- nued activity or productivity of that system (Wiener 1961). The information travels from a system through its environment and back to the system, where feedback reports the difference between the current state and the goal, and the system acts to correct differences (Wiener, 2019). This process helps to ensure stability when disturbances threaten dynamic systems, such as machines, software, organisms, and organisations (Dubberly 2015).

The contribution of cybernetics theory to the idea of feedback is that any feedback system is automatically goal-directed, where both animals and machines fall into a new, wider class of objects that is characterised by the possession of control systems. Cybernetics not only draws analogies between animals and machines, but also studies the questions of system development at an abstract level. In other words, it provides a common ground to talk about living organisms, including humans, and machines, in a language sui- table for a description of any goal-directed system. This last assumption has essentially determined the specific means and methods used in cyber- netics (Valée 1995). In a broad sense, cybernetics involves the study of how systems regulate themselves and act toward goals, based on feedback from the environment. These can be biological (maintaining body temperature), mechanical (governing the speed of an engine), social (managing a large workforce), and economic (regulating a national economy), and not just computation (Wiener 1961).

According to Walter Canon, a frontrunner researcher in biological cyber- netics, who studied mainly in mammals, and demonstrated the ability to compensate for an environmental perturbation by an internal modification rather than by an action on the environment, a feedback model that is conce- ptually explicit is called ‘self-regulation’ (Cannon cited in Parra-Luna, 2019). Control and communication play a fundamental role in cybernetics. Control is to be understood mainly as retroactive control, more precisely as nega- tive feedback for self-regulation. In these regulation problems, the purpose is to maintain an essential parameter of a dynamical system at a chosen value (Wiener 1961).

Cybernetics offers the language (both vocabulary and frameworks) that enable scientists and designers from different domains of knowledge and practice to communicate, to describe the structural similarities of systems and to recognise patterns in information flows (Dubberly 2015). This shared language is especially useful in analysing, designing, and running complex, adaptive systems.

The use of the cybernetics vocabulary (formal concepts and associated terminology) is supporting integration through interdisciplinarity and trans- disciplinarity. Its usage build bridges between different knowledge domains and allows to cross many disciplinary boundaries to create a holistic appro- ach (interdiscipline). An example is the concept of control by feedback and the associated terminology. The concept can be applied in many different domains, where engineers, anthropologists, neurologists, psychologists and

New Forms of Alliances Between Humans and Technology 89

economists (and others) are constructing “similar” models, with different domains of application and terminology, where they can share informa- tion based on common ground for modelisation. Thus, cybernetics serves to facilitate communication between discipline areas (Parra-Luna, 2019). The concepts, models and terminology of cybernetics become systematised as a set of inter-related concepts (transdiscipline). The impact of cybernetics as a transdiscipline is that it abstracts from many domain models of great gene- rality. Such models bring order to the complex relations between disciplines, and provide useful tools for ordering the complexity within disciplines (Ross 1956, cited in Parra-Luna 2019). Practitioners of cybernetics see the ubiquitous phenomena of control and communication, learning and adaptation, self- organisation and evolution, when looking at the world. Cybernetics is the science of human-machine inte- raction that employs the principles feedback, control and communication. Therefore, the theory of cybernetics can be used to propose a framework for the development of forms of human-AI alliances (Osmanoglu, 2022). The following sections present different human AI interaction approaches that find root in cybernetics.

HUMAN IN THE LOOP APPROACH New types of interactions between humans and machine learning-based AI models are grouped by researchers under the umbrella term of human-in- the-loop machine learning. Different approaches of human-in-the-loop are defined based on who is in control of the learning process. These can be identified as: active learning, in which the system remains in control; intera- ctive machine learning, in which there is a closer interaction between users and learning systems; and machine teaching, where human domain experts have control over the learning process (Munro 2020). Aside from control, humans can also be involved in the learning process in other ways. In curri- culum learning, human domain experts try to impose some structure on the examples presented to improve the learning; in explainable AI the focus is on the ability of the model to explain to humans why a given solution was chosen (Mosqueira-Rey et al. 2022).

Human-in-the-loop allows to leverage the power of the machine and human intelligence to create machine learning-based AI models. In traditi- onal approaches, humans are involved in a virtuous circle where they train, tune, and test a particular algorithm. In other words, Human-in-the-loop describes the process when the machine or computer system is unable to solve a problem, and needs human intervention, which can involve in both the training and testing stages of building an algorithm, for creating a conti- nuous feedback loop allowing the algorithm to give every time better results (Munro 2020).

INTERACTIVE MACHINE LEARNING Interactive machine learning (IML), as a type of human in the loop approach, is an iterative learning process where humans and AI are interacting based

90 Osmanoglu

on algorithms and frameworks that facilitate machine learning with the help of human interaction (Jiang et al. 2019).

Machine learning is more and more used for AI systems. The design and implementation of machine learning algorithms are developed mainly by skil- led experts. The iterative approach adopted by these experts for machine learning development process involves the collection of data, construction of the model, and assessment of the quality of the model. Potential end users, who are often domain experts for the machine learning based tool, have limited involvement in this development process. They are consulted by the experts in providing data, or giving feedback about the learned model (Amer- shi et al. 2014). With this traditional development and running process, the end users’ ability to directly affect the resulting model is limited.

Interactive machine learning addresses the question of how end users’ invo- lvement can be embedded in AI enabled tools, by defining AI subsystems that interact with humans. Caruana and colleagues brought the idea of creating a new cluster in the AI cluster that can allow interactions with end users, and they developed algorithms that enable interactive exploration of the cluste- ring space and incorporation of new clustering constraints (Caruana et al. 2006). This approach enables everyday users to interactively explore the model space and drive the system toward an intended behaviour, reducing the need for supervision by practitioners. By allowing users to train, classify, view, and correct data through a feedback mechanism, interactive machine learning empowers them to create sub systems within machine learning for their own needs and purposes (Amershi et al. 2014).

DESIGN FOR AI ENABLED TOOLS Both iterative design processes and cybernetics are based on learning and adapting to the needs of a system. In that matter, cybernetics can provide a useful framework for augmenting designers in creating human-centred inte- ractive AI-enabled products (Dubberly 2015). Feedback loops defined within the framework provided by the cybernetics theory can be used by designers in embedding them into AI-subsystems to allow interaction with users, that can enhance the AI subsystem to further learn and adapt to the user.

Researches conducted on human-in-the loop design approaches, and par- ticularly on interactive machine learning, show these share a common pri- nciple; that intelligent systems are designed with the goal to enhance or augment the human, and to evolve through human interaction. The Stanford University human-centred artificial intelligence working group reframed ‘the human-in-the-loop’ design approach as human-computer interaction (HCI) design problem, based on defining a selective inclusion of human participa- tion in the intelligent automation as input, where the result measures the efficiency of intelligent automation while accommodating human feedback and remaining pertinent (Stanford HCI, 2022). In this approach, each step that incorporates human interaction expects the system to be designed in a way that before taking the next action these steps are understood by humans, and the critical steps are determined by human agency. This design appro- ach is expected to bring transparency, while incorporating human judgment

New Forms of Alliances Between Humans and Technology 91

in effective ways. AI systems are built to support humans. The value of such systems lies not solely in efficiency or correctness, but also in human preference and agency.

According to researches, it is important to focus on the information flows existing in AI systems during the design phase, in a way to make operations more efficient and user experience more meaningful by creating opportuni- ties for learning through feedback (Martellato and Ju 2018). Knowledge of cybernetics can support these processes.

The necessity of human agency in participatory design within a cybernetics-based framework, is further enhanced by designers. Medich, sta- tes that ‘AI based tools are designed to extend the human ability to think, and cybernetic models are created with an emphasis on the human to maintain the system and encode the exchanged information into computer language’. He then follows that ‘the technological side of the cybernetic system tend to adapt and learn fast from the human side, except in the encoded translated format, highlighting the importance on fluency of the human in technology to achieve more powerful things together’(Medich cited in Martelaro and Ju 2018).

HUMAN – AI INTERACTIONS IN THE CONTEXT OF ADAPTIVE LEARNING Adaptive learning tools are developed by humans, however according to con- ducted interviews and researches, learners and instructors are not sufficiently involved throughout the design and running phases of these tools. Risks and challenges associated with the adoption of these tools can be addressed with a ‘human in the loop approach’ based participatory design, where the end users would interact with AI algorithms to contribute to the outcome.

Learners and instructors can be involved in participatory design in self- managing their own learning data used for training algorithms, to avoid the concerns for example at the ethical level over the usage of their data without their consent.

Extensive interactions of learners and instructors with AI algorithms can also support the integration of technology into educational processes (peda- gogy) and better adapt better the technologies to educational processes. The interactions between AI-based tools and users can also help to develop new skills, to address the skill gap issue related to implementation at scale of ada- ptive learning tools. These hypotheses require to be supported by further researches.

Through the participatory design approach based on cybernetics, more agency is delivered to learners and instructors, at the different stages of adaptive learning tools’ lifecycle (design, development, set up and run- ning). Learners and instructors, by providing constant feedback, are expected to improve the capacity and the effectiveness of adaptive learning tools, consequently the educational system.

CONCLUSION AND FUTURE WORK Different human-AI interaction approaches that find root in cybernetics have been presented.

92 Osmanoglu

The theory of cybernetics, as the science of human-machine interactions through feedback, is proposed as a framework to develop further human-AI alliance approaches.

This paper analysed whether a participatory design, focussed on human- AI interactions, can help to address risks and challenges related to AI-based tools adoption, particularly for adaptive learning tools.

The importance of human agency, as the involvement of learners and instructors, in the context of participatory design, to enhance the design and running of AI-based learning tools, has been highlighted. The involvement of users can be overcoming risks and challenges related to the adoption of these tools, by in addressing issues related to ethics, as to the usage of data, integrating AI and educational processes, and overcoming skills shortage for implementation at scale. However, these hypotheses remain to be tested as part of future researches.

This research will be further followed up with extensive research on cyber- netics theory to develop new human-AI alliances taking into consideration new developments on AI.

REFERENCES

Amershi, S., Cakmak, M., Knox, W. B., Kulesza T.: Power to the people: the role of humans in interactive. Mach. Learn. AI Mag. (2014). Anderson, C. L. Borthwick, A.: Results of separate and integrated technology instru-

ction in preservice training. ERIC Reproduction Document # IR021919, p. 14 (2002). Ashby, William Ross An introduction to cybernetics. by: Ashby, William Ross.

Topics: Cybernetics. Publisher: New York, J. Wiley. (1956). Brinton, C. G., Rill, R., Ha, S., Chiang, M., Smith, R., Ju, W.: Individualization for education at scale: MIIC design and preliminary evaluation. IEEE Trans. Learn. Technol. 8(1) (2015). Brockman, J.: Possible Minds: 25 Ways of Looking at AI. Penguin Press, United

States (2019). Dubberly, H., Pangaro, P.: Cybernetics and design: conversations for action. Cybern.

Hum. Know. 22(é-3), 73–82 (2015). EDUCAUSE Horizon Report: Adaptive learning: understanding its progress and potential. In: EDUCAUSE Horizon Report 2019: Higher Education Edition, pp. 34–35 (2019). https://library.educause.edu//media/files/library/ 2019/4/2019horizonreport.pdf?la=en&hash=C8E8D444AF372E705FA1BF D4FF0DD4CC6F0FDD1 Emergen Research: Adaptive Learning Market Size (2021). https://www.emergenres earch.com/press-release/global-adaptive-learning-market Glanville, R.: Researching design and designing research. Des. Issues 15(2), 80–91

(1999). Green, K.: Campus computing 2018: the 29th national survey of computing and information technology in American higher education (2018). https://www.camp uscomputing.net/content/2018/10/31/the-2018-campus-computing-survey Jiang L., Liu S., Chen C., Recent research advances on interactive machine learning,

Journal of visualisation, volume 22, pages 401–417 (2019). Kaplan, J.: Artificial intelligence: think again. Commun. ACM 60(1), 36–38 (2016).

https://doi.org/10.1145/2950039 Lee, J., Park, O. C.: Adaptive instructional systems (2007). https://doi.org/10.4324/

9780203880869. CH37. Corpus ID: 415533 Published 22 December 2007.

New Forms of Alliances Between Humans and Technology 93

Lee, M.: Understanding perception of algorithmic decisions: fairness, trust, and emotion in response to algorithmic management. Big Data Soc. 5(2018), 2053951718756684 (2018). Martelaro, N., Ju, W.: Cybernetics and the design of the user experience of AI

systems. InteractionsXXV (6), 38 (2018). https://interactions.acm.org/ Mirata, V., Hirt, F., Bergamin, P., van derWesthuizen, C.: Challenges and contexts in establishing adaptive learning in higher education: findings from a Delphi study. Int. J. Educ. Technol. High. Educ. 17(1), 1–25 (2020). https://doi.org/10.1186/ s41239-020-00209-y Munro R (2020) Human-in-the-loop machine learning. Manning Publications,

Shelter IslandReturn to ref 2020 in article. Norman, D.: The Design of Future Things. Basic Books, New York (2007) O’Sullivan, P.: APLU adaptive courseware grant, a case study: implementation at the University of Mississippi. Current Issues Emerg. ELearning 5(1), 45–61 (2018). Okojie, M. C. P. O., Olinzock, A. A., Okojie-Boulder, T. C.: Technology Training Dilemma: A Diagnostic Approach (2005). https://scholar.lib.vt.edu/ejournals/JOTS/v32/v32n2/okojie.html#Okojie Osmanoglu, B.: AIS Adoption Challenges: The Role of Interactions Between Humans

and Artificial Intelligence inAdoption of Adaptive Learning, Springer Nature Switzerland AG, HCI International (2022). Oxman, S., Wong, W.: White paper: adaptive learning systems. Integrated Education Solutions, Snapwiz 1 (2014). Parra-Luna, Systems Science and Cybernetics – Vol III - Cybernetics and the Integra-

tion of Knowledge, Encyclopedia of Life Support Systems, UNESCO (2019). Peng, H., Ma, S., Spector, J. M.: Personalized adaptive learning: an emerging pedagogical approach enabled by a smart learning environment. Smart Learn. Environ. 6 (2019). Article number: 9. Pierson, M.: Technology integration practice as a function of pedagogical experts.

J. Res. Comput. Educa. 33(5) (2001) The New York Times–New Mexico Sues Google Over Childeren’s Privacy Violations, February 2020. https://www.nytime s.com/2020/02/20/technology/new-mexico-google-lawsuit.html Vallée R. (1995). Cognition et Système. Essai d’Epistémo-praxéologie. Limonest, France: l’Interdisciplinaire. [Theoretical considerations about the way a cybernetic system perceives, decides, and acts, with views on epistemology and praxiology]. Venkat, N. G.: Cognitive analytics driven personalized learning. Educ. Technol.

(2017). Weber, N.: Adaptive learning: understanding its progress and potential. In: EDUCAUSE Horizon Report 2019: Higher Education Edition, pp. 34–35 (2019). https://library.educause.edu/media/files/library/2019/4/2019horizonrepo rt.pdf?la=en&hash=C8E8D444AF372E705FA1B F9D4FF0DD4CC6F0FDD1 Wiener, N.: (reissue of the 1961 edition) Cybernetics or Control and Communication in the Animal and the Machine. The MIT Press, Cambridge (2019). Wiener, N.: Cybernetics or Control and Communication in the Animal and Machine,

vol. 25. MIT Press (1961). Yang, T. C, Hwang, G., Yang, S.: Development of an adaptive learning system with multiple perspectives based on students’ learning styles and cognitive styles. Educ. Technol. Soc. 16(4), 185–200 (2013). Zhang, K., Aslan, B.: AI technologies for education: recent research&future directi-

ons, computers and education. Artif. Intell. 2, 1000025 (2021).

Cover of the proceedings volume Training, Education, and Learning Sciences

Explore the complete digital volume

Training, Education, and Learning Sciences

Read this individual research paper free through the original publisher, or explore the full digital proceedings book.

ISBN 978-1-958651-85-8

Explore this ISBN in the AHFE Bookstore Listed in the July 2026 bookstore catalogue; search by ISBN to confirm current availability · Digital book purchase optional

Explore Related Research

Related research 8 matches

More papers from this volume

← Back to Research ExplorerBrowse the full library →