Repulsion in controversial debate drives public opinion into fifty-fifty stalemate,Phys. Rev. E 100 042307 (2019) [arXiv]
Opinion formation is a process with strong implications for public policy. In controversial debates with large consequences, the public opinion is often trapped in a fifty-fifty stalemate, jeopardizing broadly accepted political decisions. Emergent effects from millions of private discussions make it hard to understand or influence this kind of opinion dynamics. Here we demonstrate that repulsion from opinions favors fifty-fifty stalemates. We study a voter model where agents can have two opinions or an undecided state in between, and where we allow for repulsion of opinions and for doubt: in pairwise discussions, undecided agents can be not only convinced, but also repelled from the opinion expressed by another agent, and decided agents may return to the undecided state. As a result, we observe that in controversial debates the frequencies of both opinions equalize. This voter model attractor reproduces the phenomenology of repeated Brexit poll data.

​ ​​  

 

The importance of antipersistence for traffic jams, Europhysics Letters 118 38005(2017) [arXiv], Deutschlandfunk Interview
Many motorways are running above their capacity and therefore suffer congested traffic. The traffic breakdown from free flow becomes increasingly likely around a critical flow, a critical number of vehicles per minute. Here we discuss congestion durations which are distributed with a power law over three decades, from minutes to hours. The traffic flow shows large fluctuations on short time scales which quickly trend back to the mean value, they are antipersistent. Consequently, it exceeds the critical flow for time spans which are power law distributed. This implies a large number of short-lasting traffic jams with a large risk for rear-end collisions. 

​

Hidden Connectivity in Networks with Vulnerable Classes of Nodes, Phys. Rev. X 6 041022 (2016) [arXiv]
In many networks, different sets of nodes are vulnerable to different failures or adversaries. To utilize redundant paths for avoiding vulnerable node classes in advance, we developed color avoiding percolation, in colaboration with Michael M. Danziger in the group of Shlomo Havlin and Vinko Zlatic at Rudjer Boscovic institute in Croatia. We found a fast algorithm for the largest cluster of nodes which can communicate avoiding all vulnerable classes, and applied it to the network of internet servers. Further we developed a theory for random network ensembles and compared it to numerical results.

​​
​

Econophysics of adaptive power markets: When a market does not dampen fluctuations but amplifies them, Phys. Rev. E 92 012815 (2015), discussed in Nature 525 (News and Views) 190-191 (2015) 
Let's assume a hypothetic market situation: Electricity markets with time-dependent consumer prices. Would this help to dampen fluctuations due to renewables? An artificial market, where many consumers perform time-flexible consumption (e.g. washing) due to individual stochastic decisions, shows extreme synchronization of demand at low prices.

​ ​

Spontaneous centralization of control in a network of company ownerships, PLoS ONE 8 e80303 (2013) [arXiv]
To understand the global network of corporate control, we developed an adaptive network model. The results imply that highly centralized control can emerge without it being an explicit goal of the most powerful companies, due to rich-get-richer effects.

​ ​​ ​​