How can bots be used meaningfully in university teaching or in companies? Can artificial intelligence be sustainable? These and many other questions were addressed by four professors of the Einstein Center Digital Future (ECDF) at the Pairing Research Talks. In 10 minutes, they first presented their own research. Then they discussed with moderator Prof. Dr. Gesche Joost and the audience.
//10 minutes Whether the next journey, the effectiveness of vitamin preparations or the causes of climate change: searching the Internet for information on such topics is part of everyday life for many people. But what about evaluating and using online information? According to Elisabeth Mayweg, there are still deficits among pupils in this area of media competence or digitalization-related competence. The ECDF professor for "Digital Knowledge Management in Higher Education" would like to take countermeasures here. "Cooperative learning and communication are beneficial for the acquisition of knowledge and for the development of competences such as critical reflection," she said. In her current research, she is investigating how the learning environment can be supported by technologies. Digital moderators could be conceivable here.
//10 minutesChristian Meske examines how the use of artificial intelligence (AI) affects the workplace. The ECDF professor for "Digital Transformation and Strategic Information Management" said that AI can help support decisions in areas such as health, finance or manufacturing. "There is a danger, however, that AI predictions are based on poor training data, learning something wrong or undesirable and not knowing what caused it," he said. As a simple example, he cited an image recognition system that had repeatedly identified a Husky as a wolf. The reason: Almost all of the wolf images contained in the training data set contained snow – as did the photo of the husky. Explainable Artificial Intelligence (XAI) is therefore exciting. This should make it comprehensible how, for example, deep learning models arrived at their results. "XIA can thereby create trust," he emphasized.
//talk In the discussion with Gesche Joost, the two researchers reported on their planned project. "In spring, we discovered that we were both investigating about social bots – but in different application areas," reported Christian Meske. His research focusses how social bots can support the start of new employees in companies (onboarding), e.g. in order to make contact more easily or to obtain initial information. Elisabeth Mayweg looks at the learning area at universities. "Critical questioning is often perceived as socially unpleasant. However, it is an important competence when it comes to reflecting and analyzing information. I would therefore like to know how social bots can support such communication," she says.
//10 minutesFelix Biessmann is ECDF professor for "Data Science" and deals with data quality and transparency for machine learning. In his opinion, many of the recent AI discussions (discriminatory or even racist algorithms) are not a problem of AI algorithms, but a data quality problem. "The quality of predictions through machine learning depends strongly on the quality of the data. When false statements are made, it is not the algorithm that is bad, but often the database," said Felix Biessmann. He is therefore working on automated methods for improving data quality. However, these methods alone will not eliminate discriminatory AIs. "Machine learning researchers and humanities scholars must work together to develop transparent systems," he said. This could lead to appropriate trust in machine learning.
//10 minutes To what extent does consumption change in times of digital transformation? ECDF Professor Tilman Santarius is investigating this question. "Platforms and apps have made offers such as car sharing and food sharing possible on a large scale. A few years ago, there were only a few people who took food from containers at supermarkets. Today there are hundreds of thousands of people involved in food sharing," he said. On the other hand, the mass market is growing due to digitalization. "We have drastic increases in online sales, while the stationary retail sector remains stable," Tilman Santarius said. The reasons for this are personalized advertising, the linking between social media and advertising and constant availability.
//Talk In the following discussion, the topic "Streaming services and energy consumption" attracted the attention of the audience. According to Tilman Santarius, energy consumption and emissions are rising drastically due to the high global data throughput on the Internet. In the private sector, streaming today is the biggest demand for bandwidth and generates the largest amount of data traffic on the Internet. Felix Biessmann added: "AI can also help here. If intelligent predictions of streaming behaviour could better know, how the expected stream volume will be, energy consumption could be kept lower". Tilman Santarius and Felix Biessmann want to pool their expertise in the future to develop an AI assistance system for sustainable consumption.
The Pairing Research Talks are part of the Berlin Science Week, which runs until 10 November 2019 in Berlin.