General My research combines theoretical and experimental work and deals with the relationship between Behavior, Ecology and Evolution. While I may be described shortly as a Behavioral Ecologist, I have been also collaborating with psychologists (Ido Erev, Shimon Edelman), mathematicians (Marc Feldman, Uzi Motro, Lewi Stone), and computer scientists (Joe Halpern), in an effort to understand the evolution of learning and decision making mechanisms in humans and animals. My general view of the current challenges in the field is summarized in the goals and scope of the international workshop we organized in November 2010 (Learning, decision making, and evolutionary theory), and in a recent short commentary. My ideas about cognitive evolution (see recent paperwith Joe Halpern, and section 2 below) are probably rooted in my early studies on egg and nestling recognition in cuckoo hosts (Lotem 1993, Lotem et al. 1992, 1995), where I first encountered a simple version of the general requirement for co-evolved coordination between learning rules and the timing and the distribution of data input. My past research included studies in cuckoo-host co-evolution, parent-offspring communication and nestling begging, altruism and cooperation, decision making, mate choice, and the evolution of individual and social learning.
My current research involves three main projects:
1. The social dynamics of individual and social learning and the development of individual differences Supported by the Israel Science Foundation (2015-2019), with Luc-Alain Giraldeau. Univ. of Quebec in Montreal.
Recent years have seen increasing interest in research on social learning in animals, as well as in research on the emergence of individual differences in behavior and cognition. Yet, very little is known on how social dynamics are involved in creating such behavioral and cognitive differences between individuals, and on how such dynamics shape the evolution of social learning mechanisms. Here we propose, and aim to test, two related ideas: First, that the interaction between basic reinforcement learning mechanisms and social foraging dynamics can generate individual differences and a “skill pool effect” in social groups, and second, that social dynamics, which determine the availability and reliability of information on variable rewards, may explain the relative weight given adaptively to individual versus social learning. The second idea may also explain variation is social learning abilities within and between species. Our research will use a well-established set-up to carry out experiments with socially foraging house sparrows, and will also combine theoretical work. The current proposal was largely inspired by recent findings in our laboratory that converged with Luc-Alain Giraldeau’s “skill pool effect” theory (see also letter of cooperation), and with experimental results showing that social learning in our sparrows is strongly affected by competing information from self-experience. Our project will have three parts. First, a set of experiments is designed to determine which of two well-known types of reinforcement learning models better fits the sparrows’ behavior. In one type of models (the Relative Payoff Sum type) future choices are strongly affected by the frequency of current use, reinforcing arbitrary preferences and potentially contributing to the skill pool effect, while in the second class of models (of the Linear Operator type) such an effect is much weaker. After identifying the type of models most suitable to our system, a second set of experiments will study the skill pool effect directly: We will first train sparrows to prefer different food-related cues or to specialize on different learning tasks, and then introduce them, in varying proportions, to naïve groups that can learn both tasks. This would allow us to clarify the role of frequency-dependent learning dynamics in generating individual differences. We will run such experiments with three specialization types: color preference, learning from colors versus learning from patterns, and learning multiple cues with high or low resolution. Finally, by testing groups with different compositions of specialists or generalist individuals under similar conditions, we will assess the benefit of the skill-pool effect. The third part of our project will focus on how social dynamics influence social learning. We will first confirm preliminary findings, showing that in contrast to previous results, sparrows may actually give more weight to social information if this information is collected actively and continuously (hence providing information on both successes and failures). A second experiment is designed to generate conditions that induce maladaptive social learning, demonstrating that the adaptive balance in relying on social versus individual learning may be determined by social dynamics and reward distribution. This idea will then be explored theoretically and more extensively through agent-based computer simulations, aiming at advancing existing theory of social learning mechanisms and their evolution.
2. Learning structure in time and space: from animal foraging to language acquisition (With S. Edelman, Psychology, Cornell Univ. J.Y. Halpern, Computer Sciences, Cornell Univ., and Oren Kolodny)
The ability to learn structure in time and space and to represent it in the brain is critical for most vertebrates. The basic principles behind this ability and its evolutionary origins are poorly understood. Statistical learning methods seek patterns in the stream of data (e.g., visual objects that recur from one scene to another), from which they attempt to infer structured representations that afford generalization (e.g., a generative model for scenes that can predict the appearance of objects in novel situations). How can the search for statistical regularities be made to focus on ecologically relevant structures and avoid being overwhelmed by spurious patterns? Answers to this question are starting to emerge through a convergence of findings from several disciplines, including naturalistic observations, behavioral experiments, neurobiological studies, and computational analyses and simulations.
For the past few years I have been working with Professor J. Halpern (Computer Sciences, Cornell University) on a theoretical model for learning and cognitive development and its implications for autism (Lotem & Halpern 2008). More recently, we joined forces with a group of psychologists and proposed that a small set of principles are at work in situations that involve learning of structure from patterns of experience (Goldstein et al., 2010, Trends in Cognitive Sciences 14: 249-258). These principles include temporally and/or spatially localized alignment and comparison of data streams and the use of contingent behavioral outcomes and social interaction to boost the statistical significance of the cues to structure. A central idea in our approach is the required compatibility between the learning rules used and the attentional and motivational mechanisms directing them to process the relevant data (data-acquisition mechanisms; see Lotem and Halpern 2013, In press). We suggest that by co-evolving to handle the natural distribution of data in the animal's environment, learning and data-acquisition mechanisms are tuned jointly so as to facilitate effective learning using relatively little memory and computation. We test experimentally the predictions of this theory in social birds (sparrows) and explore its broader implications (to animal learning and human language acquisition) in computational simulations (with O. Kolodny and S. Edelman - see Kolodny et al. 2014, 2015a, 2015b & 2015c).
3. Individual decision making in a social context and the evolution of social learning mechanisms: experiments with house sparrows facing risky payoffs Supported by the Israel Science Foundation (2011-2015)
While there has been extensive research on the evolution of individual decision making under risk and, independently, on the evolution of social learning, the interaction between these two processes has received relatively little attention. Studying this interaction may be of high importance because most decision makers live in groups, and social learners make decisions under risk (i.e. facing variable payoffs). We use a well established set-up to carry out experiments with house sparrows that forage individually, socially, and as hand-raised juveniles (with a sparrow mother model). Experiments in our laboratory demonstrated that within a social group, some sparrows can develop a preference for risky patches while others develop a preference for safe patches, and that this dichotomy is explained by stochastic differences in sampling success during the learning stage. The same experiments also suggest that sparrows learn patch characteristics (color) only during independent search and ignore this information when joining other sparrows that find food (i.e. they do not use social learning in this case). These novel findings raise a series of questions demonstrating the importance of studying learning dynamics in a social context. For example, is the observed dichotomy in risk preferences related to social structure or will it emerge and persist also in a population of solitary learners? Why not use social learning when joining successful group members? Do young sparrows that follow parents also fail to use social learning, or are social learning mechanisms tuned to be activated in some cases but not in others? More generally, is it possible that apparent paradoxes in individual decision making, such as excessive risk aversion or underweighting of rare events, make better adaptive sense when their consequences are tested while in a social group? Can benefits of social foraging compensate for sub-optimal biases in individual decision making?
We address these questions using experiments and data analysis. We run parallel tests on groups of sparrows and on individuals, clarifying the role of learning dynamics in generating individual differences in risk preferences, and examining their possible fitness consequences under different group sizes and food distributions. Experiments with hand-raised nestlings are used to study the mechanisms of social learning in sparrows and the conditions for its success. In addition, the detailed experimental data will be used for fitting alterative learning models and for testing the idea that learning rules and their parameters can be modified by experience.
My past research includes some of the following projects:
Mate choice and the evolution of phenotypic diversity: the unique sexual signals of the East Mediterranean Barn Swallow Supported by the Israeli Academy of Science and Humanities (2007-2011). (with R. J. Safran, University of Colorado at Boulder, and Y. Vortman).
Evolving differences in mate preferences and sexual signals are increasingly recognized as important factors in population divergence and speciation. However, a central question is whether the differences in mating traits among populations are arbitrary or adaptive. This question is especially intriguing when different sexual signals that are used separately by different populations are found to co-exist as multiple signals in a single population of the same species complex. The East Mediterranean Barn Swallow (of the sub-species Hirundo rustica transitiva) provides a unique opportunity to study such a case. Although it is generally similar to other well-studied populations of barn swallows, this population presents an interesting mixture of two traits that are known to be under sexual selection in other populations of this widespread species complex: elongated tail streamers and a dark ventral color which are, separately, signals of male quality in Europe and North America, respectively. Understanding the coexistence of these two traits in the small East Mediterranean population may shed light on the evolutionary processes that caused the differential use of these traits in the large, widespread populations of Europe and North America. In particular, the coexistence of these sexually dimorphic characteristics may be related to the evolution of multiple versus alternative sexual signaling strategies, to the degree of genetic heritability of these traits, and to possible gene flow from other populations.
To explore these hypotheses, we proposed a four-year study of the Israeli population of H. r. transitiva. We tested experimentally whether females prefer either both elongated tail streamers and dark ventral coloration, or only one of these traits. Following Safran et al 2005 (Science 309:2210-12) we use the females' dynamic paternity allocation before and after experimental manipulation of these two male traits, as a powerful method for testing female preferences and its effects on an important currency of evolutionary fitness: paternity. We also examine whether tail length and/or ventral coloration can signal (the same or different) male qualities, such as physiological condition, probability of survival, breeding success, and parental investment (nest building, incubation and food provisioning). The heritability of the two traits is assessed by taking nestlings of known parents to be raised to adulthood in large aviaries at the Tel-Aviv University research zoo, as well as by comparing parent-offspring data from the field (using both in-pair and extra-pair young). To explore the possibility that variation in tail length and ventral coloration can be explained by gene flow from the migratory population of H. r. rustica, we captured individuals from both species in fall and winter roosts and used molecular markers to test whether morphological similarity coincides with genetic similarity. For papers summarizing the results of this project see: Vortamn et al. 2011, 2013, 2015, Dor et al. 2012, Safran et al. 2016a, 2016b).
The evolutionary ecology of social and self learning: theory and experiments in house sparrows Supported by the US-Israel Bi-National Science Foundation (BSF), 2005-2009. With Prof. Marcus Feldman (Stanford University), and Prof. Uzi Motro (Hebrew University).
Together with Professor Uzi Motro (Hebrew University) and Professor Marcus W. Feldman (Stanford University), we initiated a research project that combines experimental and theoretical work. We attempt to clarify the genetic, ontogenetic, and behavioral mechanisms of the "searcher-follower" learning tendencies in the house sparrow Passer domesticus; and to explore the theoretical implications of such mechanisms for the evolutionary ecology of social and self learning in animal societies. Using a well established research setup of captive sparrows, we carry out a set of experiments that examine whether hand-raised juveniles can be trained to become "searchers" or "followers" through reinforcement learning (see Katsnelson et al. 2008); and whether individuals that prefer to follow others do poorly in self learning tasks (and vice versa). A cross-fostering experiment will measure the genetic heritability of "searchers" and "followers" tendencies, while a complementary field work will provide additional information on the frequency of these tendencies in natural populations. In light of this research, a rich set of mathematical models and computer simulations will investigate how possible trade-offs between self and social learning may emerge, shape, promote, or suppress the evolution of higher self and social cognitive abilities, and how these trade-offs depend on environmental changes.
Learning in birds and humans: Analysis of learning rules and cognitive biases in a biological system. Supported by the Israeli Academy of Science and Humanities (2003-2007). With Prof. Ido Erev, Industrial Engineering and Management, The Technion
Many of the problems I have been studied in relation to the evolution of nestling begging or cuckoo-host coevolution are in fact quite general. My colleague, Professor Ido Erev, a psychologist from the Department of Industrial Engineering and Management at the Technion, Haifa, has been studied somewhat similar problems in human decision making. We have been working together on a project funded by the Israeli Academy of Science and Humanities, through which we combine ideas and experience from our different fields. Both economists and evolutionary biologists attempt to explain behavior in terms of its functionality and rationality, but in both humans and animals, learning mechanisms and cognitive biases may frequently produce sub-optimal behavior and cause clear deviations from maximization. Although most evidence for such deviations comes from human studies, their biological robustness and fitness consequences might well be better studied in a more basic biological system, such as the interaction between bird nestlings and their parents (see below). Our four-year study includes experiments and computer simulations; its object is two-fold: a) To test whether learning by nestlings involves basic forms of deviations from maximization, known from human studies, and to analyze their ecological and evolutionary implications; b) To study how parent-offspring communication remains evolutionarily stable despite the fact that as a result of nestling learning parents may reinforce dishonest or counter-productive begging signals.
Parent-offspring conflict and communication in house sparrow nestlings Supported by the US-Israel Bi-National Science Foundation (BSF), with Prof. David Winkler, Cornell University.
Motivated by my interest in animal communication, and my previous work on nestling begging (Lotem 1993b, Lotem, Wagner & Balshine-Earn 1999; Lotem 1998a-d), I established a research setup of both captive and free-living house sparrows in order to study nestling begging and parent-offspring communication. The project was initiated in 1995, and was expanded during the years 1996-1999 in collaboration with Professor David Winkler (Cornell University), and with the support of the US-Israel Binational fund (BSF). The first contribution of our begging project was in providing the first experimental evidence for the role of offspring learning in parent-offspring communication (Kedar et al. 2000). The project, as well as my collaboration with Professor Winkler, has continued since then, first through my sabbatical stay at Cornell (2000-2001), and currently in the form of a new chapter in this study (see Lotem & Winkler 2004; Dor et al. 2006; Grodzinski & Lotem 2007; Grodzinski et al. 2008,2009, Dor & Lotem 2009).
The evolution of altruism and cooperation
My work on the evolution of altruism and cooperation includes a study of helping behavior in cooperatively breeding cichlids from Lake Tanganyika (with Dr. Sigal Balshine-Earn, currently at Mcmaster University), and a theoretical work withProf. Lewi Stone, and M. Fishman of Tel-Aviv University. These projects were supported by the Israeli Academy of Science and Humanities during the years 1996-1999, and their derived theoretical work is still ongoing. Our study of helping behavior in cooperatively breeding cichlids showed the existence of space segregation and competition among helpers (Werner et al. 2003). It suggests that such competition may limit the opportunities to provide help, and may be motivated by direct benefits of helping behavior. It also inspired us in developing our idea of signaling components in non-signaling behaviors (Lotem, Wagner & Balshine-Earn 1999), an idea that illustrates how Zahavi's handicap principle can work together with other theories for the evolution of altruism (such as kin selection and reciprocity), and how altruistic behaviors can also evolve as signals of quality or condition.
I formally tested some of these ideas together with L. Stone and M. Fishman. Using computer simulations and game theory, we found that, in heterogeneous populations, the tendency to help should be conditioned upon individual quality (i.e. help only when you can afford the cost of helping). This phenomenon is crucial for the evolutionary stability of cooperation (via direct and indirect reciprocity) because as a result of defection by poor phenotypes, discriminative altruism by high-quality individuals becomes the dominating stable strategy (Lotem et al. 1999, Nature 400:226-227; Fishman et al. 2001). Furthermore, we have recently shown that as a result of the emergent correlation between cooperation and individual quality in reciprocity games, signaling benefits of altruistic acts (i.e. quality advertisement) can establish a stable generosity by high quality individuals that no longer depends on the probability of future reciprocation or punishment (Lotem et al. 2003).
Mate choice and mating strategies
In addition to our recent barn swallows project (see above) I have been also involved in the study of mate choice and mating behavior through some theoretical work (Fishman et al. 2003), and as part of my work with my Student, Noam Werner, who studied mate choice in the haplochromine cichlid fish Astatotilapia flaviijosephi. Our study provides the first experimental evidence for male mate choice in a lekking species (Werner & Lotem 2003), and for sequential male mate preference in this species that strongly suggests a trade-off between present and future reproductive effort (Werner & Lotem, 2006). Considering the intense competition among lekking males, our work also suggests that male choosiness may help to solve the so called 'paradox of the lek'. This is because it can make less attractive females more available to subordinate males, thereby increasing the contribution of the latter to the population gene pool and keeping genetic variability among males at a level that justifies female choice.