CIM 2018/4 Focus

Amazingly diverse and adaptable

// Katrin Schmitt

Dr Dustin Feld, Lead and Algorithm Developer of the planned Fraunhofer start-up adiutaByte, talks about the importance and uses of algorithms.

Photo: Fraunhofer SCA/Dr. Dustin Feld

Photo: Fraunhofer SCA/adiutaByte

Illustration: Fraunhofer SCA/adiutaByte

CIM: Dr Feld, do active algorithms manipulate or help us? Dr Dustin Feld: Manipulative algorithms have indeed become part of daily life, among other things through extensive, fully automated and personalised online advertising. In many areas, however, algorithms work positively as a fixture in the background and make our everyday life much easier. For example, navigation devices save us a lot of time. Algorithms support our decisions and, in my view, provide a lot of added value. In the future, I believe that algorithms will make life easier for us in other areas, as well, without being manipulative.

Please explain the most important facts people need to know about algorithms and how they are used in simple terms.
If I want to drive from Cologne to Berlin at 7:30 am today, I will ask a navigation app for the best route right now. This is the strength of algorithms and computers – they work at a breath-taking speed. The algo rithm for computing the shortest route is complex: It has to take the road network and restrictions such as congestion and speed limits into account, resulting in many unknown factors and alternatives. The initial decision directly impacts on the further course of the journey. Algorithms are great at dealing with many options and influencing factors.

Eventually, they will find a route – such as the “shortest” or “fastest” option – in a flash. However, the decision should always be up to people when it comes to situation-dependent factors, say, which an algorithm cannot know. For me, a navigation device is the perfect example of an assistive algorithm.

What algorithms are you developing as Head of the planned Fraunhofer spin-off adiutaByte?
At adiutaByte we are developing algorithms that help nursing care managers in outpatient care, calculate daily tour plans for their fleets: Who is to nurse which patient, when and where? Factors such as speed limits, forecast traffic disruptions and preferences of patients and care-givers must be accounted for. This means creating a plan in the morning that will still work by midday, and preferably meet all needs and requirements. At present, these factors are hardly considered in planning. This leads to a plan that cannot be adhered to by some staff members or that does not meet the patient’s needs.

Our algorithmic solution offers suggestions with various options: shortest routes, balanced workload for staff members, or satisfaction of patient preferences. They are listed for choice with various key indicators, such as total route length – and the planner decides. The process can then learn from the selection what factors are important to the planner and consider them in future. In this way, the algorithm becomes an assistant. It filters the options, prepares and presents them transparently – assistive algorithms!

Where are algorithms already being used to plan events?
Algorithms control visitor flows in convention centres, both during normal operations and in exceptional situations. An exceptional situation can be a medical emergency or a panic among guests. We work with our partners on upgrading our software adiuta.PLAN to incorporate such situations. Combined with machine learning methods, we can forecast the behaviour of visitor flows and assist with an orderly evacuation.

What algorithm could an event planner use to improve the visitor experience at his events or make it more efficient?
As a regular conference-goer I would like more customised lecture suggestions, based on my selection history.

What data would be necessary for this?
Essentially, data about my lecture selection in event apps that are often provided for customising your agenda.

Can these data still be collected according to the new EU data protection regulations (GDPR)?
Yes. According to GDPR this process must be more transparent – a highly sensitive legal issue. In general, it must pointed out what personal data are to be processed and why. But for this type of recommendation you do not need much data.

What fascinates you about algorithms?
Algorithms are amazingly diverse and adaptable. My team and I are working on customised optimisation algorithms for routing, fleet management and the general placement of objects. Our developments take account of real-time information from a variety of sources. No matter how divergent these applications may sound, the mathematics behind them are very similar. This diversity and adaptability enables us to use solutions in one area to inspire other areas such as the meetings industry. This is what excites me so much about mathematical algorithms.

Thank you for the interview, Dr Feld!