Agent modelling involves considering how other agents will behave, in order to influence your own actions. In this paper, we explore the use of agent modelling in the hiddeninformation, collaborative card game Hanabi. We implement a number of rule-based agents, both from the literature and of our own devising, in addition to an Information Set-Monte Carlo Tree Search (IS-MCTS) agent. We observe poor results from IS-MCTS, so construct a new, predictor version that uses a model of the agents with which it is paired. We observe a significant improvement in game-playing strength from this agent in comparison to IS-MCTS, resulting from its consideration of what the other agents in a game would do. In addition, we create a flawed rule-based agent to highlight the predictor's capabilities with such an agent.

This paper outlines the Hanabi competition, first run at CIG 2018, and returning for COG 2019. Hanabi presents a useful domain for game agents which must function in a cooperative environment. The paper presents the results of the two tracks which formed the 2018 competition and introduces the learning track, a new track for 2019 which allows the agents to collect statistics across multiple games.

Being able to work well in a team is valued in industry and beyond. As such, many university educators strive to help their students to collaborate effectively. However, it is typically the case that more than ad-hoc experience is needed to master teamwork. Often, students need to become reflective practitioners who learn from their experiences and enact change. Self and peer evaluation can help evoke such reflection. However, the facilitating conditions for effective learning from peer evaluation during group projects in computing are not yet well-defined. This research is an initial step in identifying these conditions. In this study, students engaged in a long-term multidisciplinary software engineering project in which they produced a digital game. They completed regular exercises in which they reflected upon and wrote about their contributions to the project as well as those of their peers. Thematic analysis of 200 responses to an open-ended question about the purpose of these exercises illustrated the student perspective: giving and receiving feedback; prompting personal reflection and improvement; supporting supervision; aiding marking; informing project planning and management; coming to a shared understanding of the status and progress of the project; exploring and reshaping group dynamics; improving project outputs; providing a system to hold group members accountable; and giving a sense of safety to raise issues without repercussion. Giving consideration to these differing perceptions will help educators to address concerns about group projects and lay the foundations for a model of effective learning from peer evaluation during student collaborations.

Abstract-This paper highlights an experiment to see how standard Monte Carlo Tree Search handles simple co-operative problems with no prior or provided knowledge. These problems are formed from a simple grid world that has a set of goals, doors and buttons as well as walls that cannot be walked through. Two agents have to reach every goal present on the map. For a door to be open, an agent must be present on at least one of the buttons that is linked to it. When laid out correctly, the world requires each agent to do certain things at certain times in order to achieve the goal. With no modification to allow communication between the two agents, Monte Carlo Tress Search performs well and very "purposefully" when given enough computational time. I. INTRODUCTIONThe research problem studied in this paper consists of how General Game Playing (GGP) agents perform when trying to solve a simple co-operative problem without co-operative abilities, with a focus on Monte Carlo Tree Search (MCTS). GGP is the field of writing Artificial Intelligence (AI) agents that can play a multitude of games without being written specifically for each one individually [1]. GGP in real time video games has a popular competition [2] run frequently.Games that feature co-operation of some form between human players and AI agents are commonplace. Most however feature very limited forms of co-operation that are typically scripted such as in most First-Person Shooter (FPS) games. Typically FPS games give the mere impression of cooperation, though any player that looks carefully at it will see the tell tale signs of scripting. Where FPS games typically excel at co-operation is in online modes that enable teams of humans to play against each other. Some games even provide squad structures and communication allowing direct command for the purpose of better co-ordination as in Battlefield 2142 (EA Digital Illusions CE, 2006). Real-Time Strategy (RTS) games also often have a small number of features designed to facilitate communication in a bid to facilitate co-operation. Two games that stand out for co-operation are Rise of Nations (Big Huge Games, 2003) and Empire Earth II (Mad Doc Software, 2005). Rise of Nations allowed a human and AI player to operate the same set of units and buildings, though no communication was possible at all. This allowed a form of co-operation but the AI operated to its own agenda. Empire Earth 2 allowed for humans and AI agents to co-operate by letting plans be drawn up between them that could also be followed by both the human and AI agent. These allowed a All authors are with the

No abstract

Collaborative projects are commonplace in computing education. They typically enable students to gain experience building software in teams, equipping them with the teamwork skills they need to be competitive in the labour market. However, students often need encouragement to reflect upon and synthesise their experience to attain the most learning. Peer evaluation offers one such approach, but the conditions which facilitate effective peer evaluation have not yet been established. This paper seeks to provide insight into student experiences with peer evaluation. It builds upon prior qualitative work, analysing quantitative data collected through a questionnaire taken by undergraduate students on a collaborate digital game development module. An exploratory factor analysis identifies seven dimensions of variance in the student experience: perceived impact; arbitrary influence; inconsistency; team cohesiveness; assessment pressure; ease and professionalism. Correlation analysis suggests some factors such as arbitrary influence, team cohesion, assessment pressure, and professionalism are associated with attained learning, whilst factors such as inconsistency and onerousness are not. This informs the development of a conceptual framework, suggesting focuses which facilitate effective peer evaluation. Expanding this conceptual framework and validating it across different demographics, contexts, and project types are suggested as avenues for further investigation. CCS CONCEPTS• Social and professional topics → Student assessment; • Applied computing → Collaborative learning; • Software and its engineering → Programming teams.

In this paper, we propose a social negotiation system in which agents can communicate and interact with each other socially throughout a Sheriff of Nottingham game. We address issues with the number of options available while negotiating, particularly when bluffing is involved. Experiments are proposed that would allow us to validate how closely this framework mirrors real social interaction in the game, and the possibility of generalising multi-agent negotiation beyond this framework is raised.

No abstract