AIXPERIMENTATIONLAB — Augmented Intelligence Experimentation Laboratory
Towards an institutionalized format for the design, development, use and diffusion of human-centred artificial intelligence applications
How AI addresses the challenges of tomorrow’s industry
Artificial intelligence (AI) is considered a groundbreaking technology in many fields due to its potential to reproduce or exceed the cognitive performance of humans across a variety of applications. However, use cases so far have shown that a fully automated use of AI alone is not sufficient for efficient decision-making in real-world scenarios. For this reason, a more inclusive paradigm, commonly known as augmented intelligence, has emerged alongside the adoption of automated systems. It is based on the assumption that human and machine intelligence complement each other positively. As AI-powered technologies and processes increasingly find their way into a range of industries, the relevance of designing and developing augmenting systems for complex tasks has gained importance for their widespread adoption.
“Our intelligence is what makes us human, and AI is an extension of that quality.” — Yann LeCun, A.M. Turing Award 2018 recipient
Augmented intelligence aims at synergistically complement human and machine intelligence capabilities. While AI algorithms are able to analyze and discover hidden patters in large amounts of data, humans can add context-sensitive information to enhance decision-making. Algorithms find solutions faster and more efficiently, but often only with limited certainty. To minimize this remaining uncertainty, the human counterpart provides its expertise or uses its intuition, spontaneity, and heuristics. Accordingly, the final decision-making and the resulting actions remain entirely under human control. The human in turn “feeds” the AI with his expertise and knowledge gained from experience in order to improve the algorithms for the next case. The principle of “human-in-the-loop” helps to combine the relative advantages of human and machine intelligence. As a result, not only does immediate decision-making becomes more efficient and effective, but also mutual learning takes place from the interaction between humans and machines (Figure 1).
The described type of interaction is only successful if humans and AI have a common language or a common dialog system, which in turn is implemented via a suitable interface. In this context, the human must be involved in the flow of the AI application without increasing the cognitive workload. Overall, this requires a human-centered design and development approach as well as a systematic and participative implementation process. Therefore, companies must ensure that employees are involved in and actively support the implementation of AI-based systems. Only in this way, the intended potential benefits can be realized in operational use.
Three essential design and task fields of augmented intelligence
The implementation of AI applications in the sense of augmented intelligence manifests itself in three main design and task fields: intelligence, augmentation, and transformation (Figure 2)
Intelligence: The Intelligence field encompasses the development of an AI model. Depending on the underlying use case, various statistical methods, learning algorithms, and procedures are applied for this purpose. These can be differentiated, for instance, according to their problem definition: predictive algorithms for inferring variables and descriptive algorithms for pattern recognition. Depending on the use case and the chosen algorithm, different data requirements apply.
Augmentation: Within the design and task field, the interface between humans and AI is defined. This includes aspects such as ensuring and increasing comprehensibility, defining the division of functions between humans and machines, and designing the interface, e.g., according to the human-centered design approach of DIN EN ISO 9241–210.
Transformation: For a successful transformation, i.e., a profitable introduction and use of the AI application for all participants, it is necessary to adjust the operational processes in order to facilitate communication, cooperation, and social integration. A continuous and participatory introduction process involving the employees as users is of central importance.
Augmented Intelligence Framework
Real-world AI applications involve the solution of a multitude of different tasks. These range from optimization and classification to prediction and can be accomplished individually and separately by powerful algorithms. In order to be able to simultaneously map the variability in the everyday work of employees and the use of these highly specialized algorithms, a generic framework for the development for augmented intelligence applications was developed at the Laboratory for Machine Tools and Production Engineering (WZL) of RWTH Aachen University (Figure 3).
The framework consists of several conceptual layers to allow both the interaction with collaborators, as well as the task specification and algorithm manipulation by the developers. The interface layer enables interaction with the user via defined communication channels such as e-mail, smart speakers, or graphical program interfaces. Data input from the interface layer is preprocessed, e.g., by correcting spelling mistakes (e-mail channel) or standardizing terms (program interface channel). A routing layer interprets the incoming information flows, e.g., by analyzing keywords and assigning them to the respective communication channel. The data are then forwarded to use-case-specific algorithms in the algorithmic layer. At the same time, the interpreter converts the frequently coded results of the algorithmic layer into a human-readable form by transforming classifications categorized by unique number keys into plain text. Finally, a dedicated data layer decouples the user layer from the developer layer in order to facilitate iterative application development. The data layer contains company-specific training data or data for necessary support points and structures of the algorithms. Hence, developers can process a copy of the data layer and subsequently replace and reintegrate it without friction.
The described structure enables the representation of work situations using a set of standardized algorithms that have been tested in research for different types of applications. Classification, also called pattern recognition or discrimination, assigns a dedicated class to each data point based on a classification rule. A typical use case would be to classify defect components through image recognition. On the other hand, cluster analysis can be used to explore and derive patterns from data, e.g., to divide customer data into different customer segments. For continuous data, regression algorithms can be used for prediction or interpolation, such as in tool life prediction based on machine data. In this context, the framework strives for a trade-off between acceptable accuracy of standardized and well-tested algorithms and estimable technical feasibility.
Application in the research project AIXPERIMENTATIONLAB
The presented Augmented Intelligence Framework is being tested within the research project AIXPERIMENTATIONLAB* using selected use cases from medium-sized companies with a focus on customer service. The project follows a strongly participation-oriented approach by actively involving future users already in the development phase. In this way, the user requirements, wishes, and concerns are considered from the beginning into the layer design, particularly the routing and interface layers. This ensures a high degree of usability and acceptance, which in turn reduces the initial training requirements and increases effectiveness and efficiency in the actual application. Further information and updates on the research project will follow shortly on this channel. Stay tuned!
*AIXPERIMENTATIONLAB stands for Augmented Intelligence Experimentation Laboratory — the research project runs until the end of 2023 and is funded by the Federal Ministry of Labour and Social Affairs (BMAS) as part of the funding programme “Future-oriented companies and administrations in the digital transformation (room for learning and experimental AI)” — EXP.01.00016.20.
For more information please visit the project’s official website (not optimized for mobile, in German only).
References
Bitkom (2017): Künstliche Intelligenz. Wirtschaftliche Bedeutung, gesellschaftliche Herausforderungen, menschliche Verantwortung. O. V., Berlin.
Kirste, M.: Augmented Intelligence — Wie Menschen mit KI zusammen arbeiten. In: Wittphal, V. (Hrsg.): Künstliche Intelligenz. itt-Themenband. Springer Vieweg, Berlin, 2018, S. 58–71.
