newsonline24

Optimal strategy sought

One important similarity: Reward and punishment can guide both rodents and humans in decision-making: Praise and recognition are much more desirable than rejection and rebuke. Professor Israel (Eli) Nelken from HUJI has investigated the strategies rats use for foraging.

To this end, the brain researcher examined the behavior of his four-legged protégés in a large arena (1.6 meter in diameter) called the RIFF (Rat Interactive Foraging Facility), using with audio and video recordings for tracking animal behavior at tens frames per second, and electrodes that record the activity of the nerve cells. Nelken points this right out at the beginning of his online lecture: "The amount of data is so large that we don’t fully understand everything, but I will explain how we approach this problem."

Practice makes perfect

The training area is quite simple: At several stations, the rodents can receive rewards in the form of food or water or be punished with a puff of air if they make the wrong choice. Short melodies indicate the availability of reward, while the English sentence "Don’t go there" indicate punishment. The animals learned successful strategies within a very short time: "On the first day they achieved two rewards per minute, on the second day it was already seven," explains the scientist.

The mathematical framework of Markov decision processes (MDPs) is used here. In an MDP, the best decisions are made in the moment: "It is possible to optimize the strategy by making the decisions based only on the current state, without taking into account how the rat got there", explains Nelken. Following the ideas of his long-time collaborator, the former Professor of Computer Science at HUJI: Naftali Tishby. Nelken shows that Rats use behavioral strategies that are both good and simple.

Language models as experts for corporate relationships

Following these strategies is just as important for rodents as it is for us humans. In the business world, for example, companies are often faced with essential questions such as: Partner or competitor? Customer or supplier or competitor? Understanding the relationship between companies is important in several respects, explained Sebastian Müller, Professor of Finance and Director of the Center for Digital Transformation at the TUM Campus Heilbronn, in his lecture on "Global Business Networks": Whether it’s mergers and acquisitions, the search for potential customers and suppliers or an appropriate pricing policy – every company needs to know its competitors, partners, customers and suppliers, he said.

Sebastian Müller and his doctoral student Christian Breitung followed on from a study that investigated how similar business descriptions in annual reports are: The higher the similarities, the more similar the companies and the greater the likelihood that they are competitors. In the second step, Müller and his team had a business description created by AI-based language models such as GPT-3 in order to construct global company networks. These allow various conclusions to be drawn: for example, that the shares of a company belonging to a previously successful network are highly likely to continue to perform well in the future. In addition, the language model can be refined to such an extent that a distinction can be made between competitors and customer-supplier relationships based on the text alone.

When silence is only silver

Game theory also focuses on decision-making. David Wuttke, Assistant Professor of Supply Chain Management at TUM Campus Heilbronn, made this clear in his presentation entitled "Decision Making with Game Theory." Game theory can be used to draw a number of conclusions about supply chains, such as the so-called prisoner’s dilemma. In this thought experiment, two imaginary suspects have to decide whether to cooperate with each other – i.e. to keep quiet about their crime and both be moderately punished – or whether to pursue their own advantage at the expense of the other – i.e. to betray the accomplice in order to be pardoned themselves. It turns out: "No matter what the other person does, it is always better for me to talk than to remain silent," explained Wuttke.

Equilibrium – i.e. a situation in which neither party can improve its own result – is therefore not always optimal. Applied to supply chains, this means that if all the companies involved maximize their profits, equilibrium does not necessarily lead to the optimum result –cooperation is important.

Even the game of "rock, paper, scissors" can be used to draw conclusions for corporate relationships: As with the game, it can also be advantageous in business life to deliberately use randomizing as a strategy from the outset. Otherwise you could be too predictable and be exploited by the competition.

The power of chance

So what did the participants take home with them? Rats are always hungry, AI is becoming an expert in corporate relations and business decisions are best made during a game of "rock, paper, scissors"? Host Prof. Helmut Krcmar, Founding Dean and President’s Representative for the Overall Development of the TUM Heilbronn Campus, succinctly summed up the actual knowledge gained: "The aim of the webinar was to provide different perspectives on the decision-making process. We learned that individual behavior influences decisions just as much as the given framework conditions. In addition, decision-making processes change when there are several players or chance comes into play."