Interplay of technical functionalities and human skills
For the world of work, the question arises: How can human-machine interaction (HMI) be designed in a “human-friendly” manner or in the interests of the employees? To strive for this is not only considered a social consensus, but spiegelt is also reflected in the political goals of the federal government's AI strategy: employees should be placed at the center of digital change and taken along. However, especially for the design and use of highly interactive AI systems, concepts are required so that employees are not de-qualified or even displaced, but can benefit from AI.
AI systems often work undetected in the background - e.g. to control other technical systems or at the user interface such as voice recognition in a mobile phone. However, learning systems in particular are increasingly venturing into direct interaction with people. In a highly adaptable manner, they should not only deal flexibly with the respective situation, but also learn in interaction with people and optimize the common result. Efforts are being made, for example, to use lightweight robot arms that predict the next work steps in direct close cooperation with humans and assist as a “third arm”, or software assistants who learn from observing work on the PC to independently prepare and prepare repetitive work steps take over.
It depends on the quality of the interaction
Here the focus is on the quality of the interaction. In the work context in particular, it is not enough to pay attention to usability or ergonomic, intuitive, simple use. In addition, there are questions of responsibility and decision-making or the inclusion of knowledge, experience and skills of the employees. Because these are not consumers, but bearers of a wealth of work that needs to be promoted and used appropriately.
The aim of designing highly interactive AI systems should therefore be to ensure that the functionalities of the system interact as well as possible with human skills. It is both about getting a job done and how it should be done. At work, the requirements of the work object must be reconciled with various operational goals, the interests of the employees as well as the social (e.g. legal) framework. Such a competence-based functional and framed interaction between humans and AI is assumed here as the optimum for the MMI.
The idea of complementarity between humans and AI
This is based on a consideration that should be proposed as a guideline in the change from working with AI: The idea of complementarity between humans and technology. People and technology can continue to complement each other in many places - especially in direct interaction. For this, however, it is crucial to recognize the diversity of competencies or functions of both sides, to bring them together fruitfully and to promote their development.
Even if it is important that technical interfaces are "humanized" in the sense that they adapt to the natural forms of human interaction (e.g. language, gestures, touch), with a view to a sustainable division of labor, mutual benefit must be about opportunities to expand the mutual complement. This requires a systematic examination of the differences between people and technology and in particular also of the deficits and the need for additional technical solutions. Where is AI not enough? Why do people need it?
Co-evolution of humans and AI
AI, which serves people and society, should not be restricted in its functionality, but at the same time it must be developed in such a way that it does not narrowly focus on imitating and replacing human activity, but instead searches systematically and dynamically for how technology and Human beings - in the sense of a "co-evolution" - can be promoted and advanced at the same time in their development. The driving force behind technology development should not be imitation and trumping people, but the search for a higher-quality complementarity relationship between people and technology. The best results can also be expected from this.
This applies all the more if, in the course of the introduction of AI at work, activities and job profiles are re-tailored, or central processes are automated. Because even around automated areas (upstream, parallel and downstream) there are always new activities and new opportunities for supplementary relationships - and this at all qualification and work levels.
The focus of complementarity is human-machine interaction. Their sustainable human-friendly design at AI can therefore be understood as a learning field for this thinking.
Interactive AI in concrete use in medium-sized companies
Here is an example of how interactive, adaptable AI can be used in assembly in just a few years - one of the application scenarios the Learning Systems platform. The medium-sized family company Bandorf produces cable harnesses for the automotive industry. According to the different models and individual customer requirements, different cable runs and cable lengths must be put together with high accuracy. Intelligent robot tools support skilled workers when laying the cables on assembly tables, on which the cable harnesses are precisely mounted. The sensitive tools adapt flexibly to the work steps that vary from job to job. For this purpose, the employees program the self-learning robot tools themselves on-site, if necessary, by showing them the processes in which they can assist.
Safe and diverse cooperation Classic robot systems are designed, built, programmed and tested by trained specialists for specific tasks. This process typically takes several months and always causes high costs, which often only pay off in mass production. Robot tools that can be learned, on the other hand, enable more flexible partial automation. They work reliably with people - wherever they are needed. To do this, they adapt to people and their work environment - and not vice versa.
AI in action against pandemics and their consequences
Another field of application for AI is a pandemic and its consequences. Frank Kirchner, holder of the chair for robotics at the University of Bremen, scientific director of the Robotics Innovation Center (DFKI GmbH) and head of the working group “Unhealthy environments” of the Platform Learning Systems, has how this works clarifies:
From a technical point of view, intelligent robotics could already achieve a significantly greater degree of automation than is the case - and lead to major relief for the urgently needed personnel in the pandemic. One example of this is hospitals and care facilities, where intelligent robots could take the pressure off skilled workers when cleaning, distributing food, disinfecting, or pushing beds back and forth. In agriculture, intelligent field robots could do work that harvest helpers from other countries do - be it when harvesting or spreading seed on single plants such as strawberries or asparagus. If, for example, harvest helpers can no longer enter easily, the great relief potential of AI-based robots can be easily recognized. Another area is energy supply. With the use of intelligent inspection robots, important infrastructures can be operated and monitored without human personnel. This applies to oil production facilities, for example, which may have to be cleared if an infection occurs on site. But also in the field of renewable energies - for example in offshore wind farms - permanent monitoring is necessary, which currently has to be done by specialists. Intelligent automation could provide relief here - for example through (partially) autonomous underwater robots with which communication in an emergency is even possible from the home office.
Criteria for designing the MMI at AI
So how can such a human-machine interaction be designed? Various criteria are relevant for a human-friendly MMI, which specifically aim at the design of the AI systems in the work context. In the Platform Learning Systems initiated by the Federal Ministry of Education and Research (BMBF), representatives were interdisciplinary Companies, Science and unions / associations developed a catalog of criteria. The catalog comprises a total of twelve criteria that can be bundled into four clusters:
Cluster 1: Protection of the Individual
- Health and safety
- Data protection and responsible service recording
- Diversity sensitivity and freedom from discrimination
Cluster 2: trustworthiness
- Quality of the data available
- Transparency, explainability and freedom from contradictions
- Responsibility, liability and system trust
Cluster 3: reasonable division of labor
- Adequacy, relief and support
- Actor and situation control
- Adaptivity, fault tolerance and customization
Cluster 4: Favorable working conditions
- Areas of action and extensive work
- Promotion of learning and experience
- Communication, cooperation and social integration
The catalog compares common criteria of work and technology design with the special challenges of AI systems, such as explainability, attribution of responsibility, etc. With the focus on the MMI, there are also new nuances and factors that specifically affect the quality of the interaction turn off. Two such criteria will be selected and briefly outlined here:
Agency and situation control (criterion 8)
When interacting with learning AI, employees are usually confronted with a new situation: the systems are designed in such a way that they also approach or react to the employees themselves while they work in the background. They can “prompt” or “force” their users to take predetermined or, in some cases, newly created actions, so that the AI system acts as a kind of “actor” in certain situations. This means that part of the agency or situation control is shifted to the technical system. At the same time, a sufficient level of situation control is essential for the ability to act, motivation and also health of the employees.
This conflict relationship is resolved if the handing over of the agency to the technical system does not limit, but rather expand, the human ability to act - in the sense of a complementarity relationship. In order to collaboratively and complementarily “work-sharing”, the agency and corresponding transfers must also be clearly clarified. In the interaction it must be understandable who is contributing what to a common process, who is in charge of the situation control for the respective sub-process and where or how the sub-processes are connected. It is necessary to define appropriate rules and create opportunities. This is necessary for a possible attribution of responsibility and to relieve the situation of unclear risks. A transparent and influenceable agency is the linchpin for a complementary "division of labor" between people and the AI system.
Promotion of learning and experience (criterion 11)
Humans and AI systems differ in terms of acquisition, processing and storage, reproduction and retrieval, as well as the application of knowledge. Accordingly, they learn differently, but can mutually reinforce each other. Firstly, the interaction with AI systems must be designed to promote learning and experience by enabling the acquisition of knowledge and experience in the usage process. Second, the employees must be able to interactively validate and, if necessary, correct the learning content (data quality) and learning behavior (links) of the intelligent system. In this way, the accuracy and performance of the AI system can also be improved.
A mutually supportive design also increases the likelihood that people are ready to contribute their knowledge and experience to AI systems. Such a complementary approach offers great opportunities, especially for dealing with complex situations, by meaningfully integrating machine-learned content and human experience. Especially when AI systems take on far-reaching and relevant activities and work is changing, a learning and experience-promoting design of human-machine interaction is of great importance - in order to maintain knowledge, experience and skills as well as the ability to act, but also to innovate from To promote work processes.
Looking to the future of AI: what's next?
The aim is to provide an important impetus for the sustainable, people-centered and future-oriented design of human-machine interaction in artificial intelligence. The basic attitude of the search for additional options for people and AI in the MMI (and beyond) is central. Even in the event of hard "disruptive" changes and changes in the industry, the employees concerned can be "taken along" when new complementarity relationships are sought and development options are offered and work and technology are designed accordingly.
Such a perspective makes the digital transformation more tangible overall, since it offers future options for people and technology development alike and does not necessarily presuppose a digital split: in a few highly qualified people for some creative and social activities and many low-skilled people for the activities whose automation (still) ) not worth it. It contrasts the vision of the fully automated “Smart Factory” with the vision of a “Smart Empowered Factory” made up of decentralized human-technology collaboration units at all work levels. Such an organization would be more agile or flexible, could fall back on a broader set of technical functionalities and human skills, knowledge and experience and would be at least as realistic to implement - not least to prevent new "CIM ruins" of failed AI automation projects.
More knowledge - PDF download, eCourse on demand or personal advice
Offline download: Download this text as PDF - Read usage rights, Because we do not automatically submit the title of this text for privacy reasons: When buying in "interests" the title register if support is needed. After buying text exclusively Download at this URL (please save). Or for a little more directly an entire book or eCourse with this text buy, read on.
Read customer feedback and buy a book on this topic at a discount: Do you like this text and want to read more information about it? Buy the right book including this text, buy it here in two languages, as a member even with a 20 percent discount. Would you like to take a look at the book first? You can do this by previewing the book look at and then purchase on the book page.
German edition: ISBN 9783965962446
English version: ISBN 9783965962453 (Translation notice)
Your eCourse on Demand: Choose your personal eCourse on this or another desired topic, As a PDF download. Up to 30 lessons with each 4 learning task + final lesson. Please enter the title under "interests". Alternatively, we are happy to put together your course for you or offer you a personal regular eMailCourse including supervision and certificate - all further information!
Consultant packages: You want to increase your reach or address applicants as an employer? For these and other topics we offer special Consultant packages (overview) - For example, a personal phone call (price is per hour).