Group 3 Copy
0 Results
Schließen
May 8, 2019

AI in focus: Human-robot collaboration – turning tools into companions

It's the talk of the day: artificial intelligence (AI). If you want to join in the conversations, you shouldn't miss our new series "AI in focus" on this groundbreaking topic. In the second part, Michael Burmester looks into the collaboration between humans and AI. How did the advent of AI change interaction patterns? What requirements need to be met for the human-AI team to work well?

Human-computer interaction – now and then

In the beginning, there was a command-reaction paradigm: The computer awaits a user command such as a mouse click. This triggers a process or opens an app. In such a model, the user takes the initiative and is in control. This hierarchical relationship has begun to change though: Almost unnoticed, the computer is turning from a subordinate tool to a partner. When we leave work, our smartphone tells us how much traffic there is and what time we will be home. Analyzing data allows it to conclude what the users will do next and provides them with useful information. So, computers are no longer restricted to reacting to commands, they make their own suggestions.

Kikonket Teil02 Bild 01 Partner

From tools to companions

Human-robot collaboration at the workplace exemplifies the metamorphosis form a tool to a companion particularly well. Here, the term collaboration is used to describe the cooperation between humans and robots, pursuing a common target. And the road to success is, unsurprisingly, well-coordinated team activity. Robots utilize AI to physically work together with us. They carry out physical steps in the production process, such as lifting heavy objects, for us. And our digital companions can even support us in non-physical, intellectual tasks.

What makes a dream team tick?

Four requirements need to be fulfilled for the human-AI collaboration to work. An example provided to us by Stuttgart Media University (HdM) and Siemens AG helps us illustrate these general requirements and their specific significance. In the project, we simulated the behavior of a digital companion. Using previously recorded speech files and text speech synthesizers, the PC user is able to interact with the prototype via speech and screen display. The scenario: You have just started with a company and are assigned the task of planning a workshop on "Communication within teams". So far, you have little experience with such assignments. But, no worries, your companion will help you – but only as much as you like.

Common goals & acceptance

First, the human partner has to be willing to work with a robot and to accept the common goal. The way to achieving this goal is flexible, both partners can adapt their work steps individually.
Back to our scenario: You and your companion have a common goal – you are planning a workshop. The jobs and roles are clearly defined. The companion provides you with its knowledge on workshop structures and methods and creates documents for the preparation. You choose what you need.

Kikonket Teil02 Bild 02 Ziele
Kikonket Teil02 Bild 03 Vorhersehbarkeit

Mutual predictability

Another requirement is that the human partner as well as the robot need to be able to assess what the other one will do next.
In our example, you companion explains exactly what it is doing. You can watch it creating documents on the screen. At the same time, it will be there during your planning steps, asking questions to see how far you have come.

Adaptability

Requirement no. 3: Intelligent agents need to be adaptable and controllable. This enables a core ability of AI – learning. Through learning, work processes can be continuously adapted, optimizing the system without having to reprogram it.

But how does your companion adapt? You might know a moderation technique that it doesn't know. It takes up your input and memorizes it – it learns. This allows it to change, adapt and optimize its procedures.

Common ground

The last requirement is what we call "common ground". The term is used in linguistics to describe the fact that communication partners make assumptions with regard to the knowledge of the other party. That way, they can comprehend the knowledge base a dialog partner uses for reasoning.
How about you and your companion? You know exactly what abilities your companion has because it makes them transparent and, on request, shares them with you. And your companion, in turn, is aware of your status because it records your planning process.

Kikonket Teil02 Bild 04 Common Ground

Distribution of roles

Another important aspect to be considered in the definition process is the distribution of roles. Since the previously defined role of the robot influences its behavior and its actions, it is important to clearly define what role the companion is to take. Would you like it to be like a teacher who instructs you and shares its knowledge or rather like an assistant who only reacts on request?
In our example, your companion can have two different roles, depending on the task assigned to it. As an advisor, it offers its expertise and learns from your contributions, extending its knowledge base. As an assistant, it only acts when you want it to, e.g. by recording your work steps or completing a planning document.

My friend, the AI?

Recent studies show that human-AI collaboration not only works well, it is also perceived as natural. Even though machines are considered a technical object, they are also seen as a social counterpart with their own intentions. In the planning stage, roles and characteristics need to be defined in a way that ensures natural interaction while preventing fears and a lack of trust. This is key to achieving a positive user experience. But how can we get there? Stay tuned, in our next part we will talk about creating human-AI interaction for a positive user experience…

News Buzzword Burmester 935

The author

Michael Burmester, professor at the Stuttgart Media University HdM and UX expert, writes for UID.