Research Interest and Results


My main scientific interest is in biological pattern formation, individual based modeling, evolution of complex systems, and decentralized decision mechanisms. I believe that these are among the most important and general theoretical backgrounds in the fields of behavior, ecology, and evolution.

Current projects

Behavioral regulation and the organization of societies (main project)

My primary and current basic research program is about how self-organizing patterns are made up by autonomous agents interacting with each other and the environment. The model systems I have studied both experimentally and theoretically are insect societies. I have shown that several colony level phenomena can be understood only as emergent properties of the system

General context and significance of the project

  1. Social insect colonies are one of the largest and most easily observable biological entities that operate without central or hierarchical control. Like cells in developing embryo or neurons in the brain, the builders of social insect colonies respond to simple local information. New global patterns and behavior emerge from the simple rules due to individual local interactions. As Franks et al. (1992) pointed out, building behavior of social insects is a classic example of fundamental problem of biological pattern formation (see Murray 1989). How such a system works is one of the fundamental questions of biology.
  2. Behavioral ecology has collected enormous information about WHY behaving in a given way is adaptive. These functional answers proved to be extremely important in understanding the evolution of biological systems. However, questions about HOW are missing in the majority of cases. Neurobiology and physiology have heretofore been unable to provide explanations for several behavioral processes and patterns. Social insect societies have proved to be one of the most tractable and most deeply studied systems where connecting the WHY and HOW questions has proved to be fruitful (Wilson, 1971; West-Eberhard, 1996).
  3. Natural selection changes the distribution of genes within the population, and the unit of selection can be linked to individuals in the short term. However, in social insects individuals function in the context of society, where reproduction is assigned to one or few individuals, but the success of reproduction is dependent on the whole colony. To understand how a worker contributes to the fitness of nest mates and its own we first must understand how behavior at the individual level contributes to colony needs and how it is integrated into the behavior network of the whole system. To study regulatory mechanisms such as stigmergy is crucial in elucidating the evolution of societies (Karsai and Wenzel, 1998; Camazine et al. 2001).
  4. The traditional computational paradigm of robotics seems to be vulnerable in an unstructured dynamic environment. Behavior-based architectures inspired by biology may provide a reasonable alternative (Connell, 1990). Combining behavior-based architecture with biologically inspired control algorithms (such as clustering object by ants or building by wasps) seems to be a promising approach. Besides exploration, the main duty of autonomous robots is and will be to build something or rearrange objects. My goal is to examine such regulation mechanisms in a given biological system, which maybe used later in as a (part of) control algorithm in an artificial system (Karsai et al., 1996).


  1.  Using the idea of stigmergy (previous work directs and triggers new actions) for generating artifacts and analyzing natural nest architectures, I showed that several properties of the nest emerge automatically without invoking global perception, memory, information collecting and processing, or counting ability of the builder (all of them were previously assumed in related scientific literature). (See information on these studies in Karsai and Penzes, 1993; Penzes and Karsai, 1993; Karsai and Wenzel, 1995). Detailed observation on the postures of the builder wasps revealed that as the nest is developing it provides new stimuli both in number and kind, and the developing structure sometimes allows the wasp to take new building postures. This dynamic interaction between the structure and the builder drive the nest to develop in its characteristic way (Karsai and Theraulaz, 1995).
  2. With a simple model I showed that different nest shapes in paper wasps can emerge from the same basic building rule. There was no need to invoke any large change (e.g., macro mutation) in the building program. A slight change in the single tuning parameter of the low complexity rule resulted in a large and characteristic global change in the generated artifact. With strict analyses of the natural shapes of the nests of genus Polistes we showed that we were able to generate all the characteristic natural forms and only those (Karsai and Penzes, 1996, 1998).
  3.  Studying nest cell initiation using natural architectures and simple modeling revealed that, contrary to previous belief, paper wasps optimize cell arrangement and not material usage. The existence of non-optimal forms led us to analyze several rules of thumb, which might be used for the wasp to construct the natural nest forms. Our aim was to find such a rule that can produce all natural forms, all optimal forms, and also predict all non-optimal forms as a byproduct of the rule of thumb (only local information was available, no information processing of the builder was invoked). We showed that the structure itself provides enough information to govern its own quasi-optimal development (Karsai and Penzes, 2000).
  4.  In a larger scale comparative study we showed how parallel processing and reduction of the size of individuals emerge as colony size increases in order to increase the system reliability and productivity. We showed that large colony size commonly corresponds with complex colony-level performance, small body size and lower per capita productivity. Parallel processing by specialists in large colonies provides flexible and efficient colony level functioning. On the other hand, individual behavioral flexibility of jack-of-all trades workers ensures success of the small and early societies. (Karsai and Wenzel, 1998).
  5.  The dynamics of colonies show interesting fluctuations. Optimization of energy budget, different effects from workers, and voluntary cessation of egg laying have been assumed. We proved that it is not necessary to invoke any of these. The interaction of (experimentally supported) simple positive feedback mechanisms was sufficient to generate the emergent properties of this dynamics. Using data from natural colonies and fitting techniques we showed that our model predicts the observed numerical changes of different stages (pupa, eggs, larvae) and the bursts in hatching. The system reacts exactly the same way to perturbations as natural colonies (Karsai et al, 1996).
  6.  Nest construction of social wasps is an excellent model system to study division of labor and the performance of a decentralized behavioral regulation. On the basis of our own field studies (Karsai and Wenzel, 2000), and literature data we suggested a new mechanism for the regulation of construction behavior in social wasps, where a natural substance (water), which is itself also a building material, regulates the colony level performance. By experimenting a simple model system, we showed that the model predictions agree with the observational data and cover a wide range of evolutionary transitions (Karsai and Balazsi, 2002). We concluded, that instead of the chained flow of information from one task group to the next (as it was assumed in the literature) a regulatory substance (water) regulates task allocation and nest construction.
  7.  The effect of food quantity on the morphology and development of Polistes metricus were studied and the experimental results were compared with the predictions of the parental manipulation hypothesis. We observed that extra food in larval stages did not result in larger adult wasps, but their abdomen became larger and heavier, and they survived the cold test longer. We infer that these colonies produce more gynes and fewer workers than control colonies. Results of a restricted nourishment treatment did not support the differential feeding (parental manipulation) hypothesis (Karsai and Hunt, 2002).
  8. Studies on self-organization and parallel processing have become important for the applied sciences. Algorithms extracted from insect societies have been implemented into different robotic or artificial intelligence systems. This is the field where biology, artificial life, artificial intelligence and robotics have a promising and important overlap. Collaborating with a researcher with experience on robot building and programming, we made preliminary implementation of the wasp building algorithm into a simulated robot system. The duty of the robots is to pile individual boxes into a compact pile. The robots are very simple and they neither possess a global view of the pile nor do they carry out complex information collecting and calculations. We proved that decisions made on the basis of stigmergic local information are sufficient to govern the piling process near the optimal solution (Karsai, Penzes and Altenburg 1996).
  9. The regulation of task partitioning in wasp societies is a complex and adaptive trait. Allocation of the construction workforce into 4 linked task groups ensures steady construction as an emergent phenomenon. We propose that the common stomach as a communication platform in worker connectivity serves both as a temporal storage for one of the building materials (water) and also as an information source (information center) about the colony needs that relate to nest construction (Karsai and Runciman, 2009; 2011). We will also show that this system is able to dynamically adapt to perturbations of the environment and to changes in colony-level demands or population structure Karsai and Phillips, 2012: Karsai and Schmickl, 2011)

Other Projects

Ecological pattern formation



Plans for the future

  1. I am keen to continue my current multidisciplinary research, where the main problem is concerned with decentralized decision mechanisms in behavior. I consider that the individual-based models in complex systems (both proximate and ultimate questions) are a fruitful approach for this challenge. I would like to extend my plans beyond my current project studying other type of behaviors (e.g., foraging) and the mechanisms of the division of labor.
  2. My plan is to extend my approach also to new areas namely to ecology and developmental biology.

    "One challenge of developmental evolutionary biology is to demand more precision in pinpointing the actual effects on phenotype ontogeny...This requires attention to the mechanisms of regulation" (West-Eberhard, 1996 pp. 293). Social insects show remarkable phenotypic plasticity. I would like to study the effects of food quantity on the morphology, development and productivity of social insects and compare the experimental results to predictions of adaptivity models (such as the parental manipulation hypothesis).
  3. As for the ecological extension of my work, we are developing software and writing a book that explains how fundamental "biological rules" (such as selection, stabilization, catastrophes, random walk, transformation and competition) can be derived from more general scientific knowledge without invoking any information that is special only for biology. Our aim is to promote the importance of emergent patterns in biology and to provide clear examples of how complex spatio-temporal patterns emerge from simple rules and local information (free beta software:
  4. I believe that the dynamics of the colony, or in larger scale, the population dynamics of the social insects, can be understand only if we account the elementary behaviors, the developmental constraints, ecological conditions and the feedback mechanisms. This integration is my long-term purpose in this field.
  5. I have been very successful with collaborating on different projects. I have collaborated with ecologists, ethologists, evolutionary biologists, ornithologists, entomologists, behavioral ecologists, phylogenetists, molecular biologists, physical chemists, physicists and roboticists. I enjoyed these multidisciplinary collaborations and my modeling and analytical abilities were useful for my partners.

Past Projects

Ecological pattern formation

  1. I defended my Dr. Univ. degree at Jozsef Attila University, Hungary, as a community ecologist in 1990 (Thesis: Indication and sensitivity of sphecid community to habitat heteromorphy). I studied different problems of how heteromorph habitats are exploited and used by insects possessing different sensitivities and habitat ranges. This study involved a combination of collecting insects with traps, plus field studies and observations of living animals (carabid beetles, sphecid wasps and pompilid wasps). I determined the collected material to species and used multivariate techniques and different null models to show different activity and fidelity to habitat types for different groups (Karsai, 1988, 1989; Preiszner and Karsai, 1988, 1990). These studies also had local importance for the management of the Kiskunsag National Park (Hungary’s largest and most famous NP), and these studies became a part of the development of a global ecological monitoring system in Hungary (Galle et al. 1990, scientific report for the Government).
  2. Using simulations and combining multivariate analyses with calibration curves, we worked out a new method to distinguish noise and signal in the spatio-temporal rearrangement of populations of leafhoppers (Gyorffy and Karsai 1991 Animal Ecology). This new method increased the reliability for interpreting changes in the spatio-temporal distribution of animal populations. After improving the model, we successfully applied this approach to carabid beetles that moved between two different habitats (Karsai, Barta and Szilagyi 1994).
  3. Genetically manipulated plants may pose risks for natural ecosystems, if the plant or the gene escapes from the controlled agricultural field. The Hungarian National Committee for Technical Development supported an interdisciplinary team to carry out the program entitled: Risk analysis of environmental and cultivation effects using genetically manipulated plants in agriculture (detailed scientific report in Hungarian: Dudits et al., 1995). I collaborated with molecular biologists and agronomists. Our aim was to work out a system where this problem could be studied in the field and to study the distribution and spread of marker genes. The results, as we expected, varied considerably depending the plant species and the mode of agronomy involved in the experiment: one of the largest spreads was observed on Brassica napus where the marker gene was detectable at 32m from the originally genetically manipulated parent plants (Pauk et al., 1995).

Behavioral ecology: food-offspring relationship

  1. Collaborating with T. Szekely I studied the availability of the food of Kentish plover (Charadrius alexandrius) and the factors that affect the clutch-size of this bird (Szekely, Karsai and Kovacs, 1993; Szekely, Karsai and Williams 1994). We found that Collembola and Diptera were the most common available food by number out of the 14063 arthropods we identified and measured. The expectation of clutches hatching over time was positively correlated with both prey density and prey mass. These studies are well cited in the literature as an important scientific contribution, and also provided important data for the management policy of the Hungarian Conservation Biology Agency concerning this endangered species.
  2. I lead a long-term project that aims to describe the sex allocation and host choice mechanisms of a pompilid wasp (Anoplius viaticus paganus Dahlb.). This effort involved a combination of field observations and laboratory experiments. Sex allocation models predict that the mother will lay a female egg onto a large host and a male egg on a small one. However, we also would like to gain detailed data on the distribution of available hosts and their quality change in time in the field. This wasp is specialized to a very small time window of the year where some spiders reach mature size but have not yet laid their eggs. While mature male spiders are rejected as hosts for the larvae, the premature males and females were used in equal frequency. Large spiders were captured by wasps of all sizes, but small spiders only by small and medium sized wasps (Karsai and Vajda, 1991, Somogyi MS thesis, 1995, Karsai, Somogyi and Hardy in prep.).