Towards Human-like Social Multi-agents with Memetic Automaton Liang Feng, Yew-Soon Ong, Ah-Hwee Tan and Xian-Shun Chen Centre for Computational Intelligence School of Computer Engineering Nanyang Technological University Email: {FENG0039, ASYSOng, ASAHTan, CHEN0469}@ntu.edu.sg Abstract—Memetics is a new science that has attracted in- creasing attentions in the recent decades. Beyond the formalism of simple hybrids, adaptive hybrids and memetic algorithms, the notion of memetic automaton as an adaptive entity that is self-contained and uses memes as building blocks of informa- tion is recently conceptualized in the context of computational intelligence as potential tools for effective problem-solving [1]. Taking this cue, this paper embarks a study on Memetic Multi- agent system (MeM) towards human-like social agents with memetic automaton. Particularly, we introduce a potentially rich meme-inspired design and operational model, with Darwin’s theory of natural selections and Dawkins’ notion of a meme as the principal driving forces behind interactions among agents, whereby memes formed the fundamental building blocks of the agents’ mind universe. Experimental studies on a Mine Navigation Task indicates the modeling of memetic agents that resemble the natural way of human interaction can lead to greater level of adaptivity and effective problem-solving. I. I NTRODUCTION In the last decades, the new science of memetics which represents the mind-universe analogue to genetic in culture evolution has stretched across the field of biology, cognition, psychology, etc., and attracted significant attentions. Memetic computation is defined as a paradigm that uses the notion of meme(s) as units of information encoded in computational representations for the purpose of problem-solving [2]. It covers a plethora of potential rich meme-inspired computing methodologies, frameworks and operational algorithms. In Dawkins’s book entitled “The selfish Gene” [3], the term meme is defined as “the basic unit of cultural transmission via imitation”. In the context of computational intelligence, meme has been typically perceived as individual learning procedures, adaptive improvement procedures or local search operators that enhance the capability of population based search algorithms [4], [5], [6]. This integration has been established as an extension of canonical evolutionary algorithm, by the names of hybrid, adaptive hybrid or Memetic Algorithm (MA) in the literature, and used successful for solving many real world search problems, ranging from continuous optimization [4], [7], combinatorial optimization [8], [9], constrained optimiza- tion [10] to image processing [11], etc. In spite of the success that meme has enjoyed, it has remained to play much of a complimentary role in the “learning” phase of the evolutionary cycle. Thus the true nature and potential merits of memes remain yet to be fully exploited in the context of computational intelligence. Beyond the formalism of simple and adaptive hybrids in MA, Situngkir presented a structured analysis of culture by means of memetics, where meme was regarded as the smallest unit of information[12]. Heylighen et al. discussed the replication, spread and reproduction operators of memes in cultural evolution [13]. Nguyen et al. [14] studied the notion of “Universal Darwinism” and social memetics in search, and investigated on the transmission of memetic material via non-genetic means. Meuth et al. [15] demonstrated the potential of meme learning and high-order memes for more efficient problem solving while Acampora et al. [16] intro- duced memetic agents as intelligent explorers to create “in time” and personalized experiences for e-Learning. In contrast to memetic algorithms, significantly fewer studies on other manifestations of memes for effective problem solving have been explored, making it a fertile area for further research investigations. The current paper thus presents an attempt to reduce this gap. In a recent survey on the multi-facet views of memetic com- putation, “memetic automaton” is defined as an adaptive entity that is self-contained and uses memes as the building blocks of information that facilitates problem-solving [1]. Conceptual- ization of memetic automaton unleashes the significant number of potentially rich meme-inspired design, operational models, and algorithm frameworks that could form the cornerstones of memetic computation as tools for effective problem-solving. Falling back on the basic notion of meme, as the fundamental building blocks of culture evolution, this paper embarks a novel study on Memetic Multi-agent system (MeM) towards human-like social agents with memetic automaton. In the field of multi-agent system research, one of the core objectives pertaining to social learning is to benefit from the mutual interaction through collaborations [17]. Oliveira and Nunes in [17] showed that the learning process of agents can be improved by means of advice exchange among the social agents. In particular, to enhance personal problem solving capabilities, under-performing agents can seek advice (on behavior) from other friendly elite agents. Rafael et al. [18] studied different means of information sharing among the agents such as case-based reasoning, while Melo et al. [19] investigated various forms of information that may be imitated