A short overview of 5 publications

[Maciej Świechowski]

Dear Members of the Community

Since there are not so many messages going round, I would like to take the opportunity to introduce you to some of my most recent publications:

1. "The Many Challenges of Human-Like Agents in Virtual Game Environments"
- presented at AAMAS'25 (24th International Conference on Autonomous Agents and Multiagent Systems)

https://dl.acm.org/doi/10.5555/3709347.3743837
https://arxiv.org/pdf/2505.20011
https://www.cyberiada.eu/aamas25.html

This work is composed of two sections. The first is a mini-survey identifying 13 key challenges associated with either creating or assessing human-likeness in games. It begins by questioning whether human-likeness is a well-defined concept given the diversity among humans!

This is a conference paper, so the survey had to be limited in terms of how deep each topic is discussed, however, the range of topics is quite exhaustive (I think!). The survey is based on the analysis and synthesis of 54 papers.
The second part describes our empirical experiment aimed at training a machine learning model to distinguish between human players and bots.
Our intention is to utilize this detection model as part of the game development pipeline. It serves as QA/assessment in creating believable NCPs - by evaluating them based on how well they can fool the model.

2. "Combining Tree Search with Value Prediction Models Capable of Estimating Their Uncertainty"
https://ieeexplore.ieee.org/abstract/document/10964843/
https://www.cyberiada.eu/CombiningTreeSearch.htm

This paper has been accepted for publication in the IEEE Transactions on Games and is currently available in 'Early Access'. Full bibliographic information is expected by December 2025.

The work explores a simple idea of combining tree search (in this case, MCTS) with a value prediction model that is applied dynamically based on its confidence in a given state, rather than randomly or uniformly, such as at a fixed depth.

So far, this approach has been very strong whenever I used it, but a model that accurately calculates the confidence or uncertainty of its prediction is not always available.

3. "AutoGT: Automatic Generation of Game Trees for Algorithm Benchmarking"

Also accepted to IEEE Transactions on Games and currently in Early Access as well.
https://ieeexplore.ieee.org/abstract/document/10984425

The idea is to procedurally generate game trees from the top-down for benchmarking search algorithms.
Part of the generated data is the optimal value in terms of Nash equilibrium - it is procedurally propagated (according some rules) from top-down without the need of calculating it by back-propagation (which would require having the complete game-tree).
Consequently, the generator is lightweight and memory-efficient.

4. "Monte Carlo Tree Search: a review of recent modifications and applications"

https://link.springer.com/article/10.1007/s10462-022-10228-y

Unlike the aforementioned works, this survey was published in 2022. Nonetheless, it remains my most cited publication (Google Scholar shows 415 citations), and I keep recommending it, particularly the MCTS + ML section.
Nowadays, MCTS is sometimes applied in the context of LLM agents - which gives it some resurgence in interest.

5. "The History of Artificial Intelligence: From Leonardo da Vinci to Chat-GPT"

Finally - my book about the history of AI.
https://www.amazon.com/dp/B0DMMKMPXL

This work highlights what I consider to be the most pivotal milestones in the field, so I am aware that it is subjective.
I think that it might be interesting to see which milestones I have chosen.
The table of contents is available here on researchgate:
https://www.researchgate.net/publication/385771937_The_history_of_Artificial_Intelligence_From_Leonardo_da_Vinci_to_Chat-GPT

I appreciate your attention and I wish Everyone a great day,
Maciej Świechowski