iiinova — where your digital self keeps evolving

and with that, my job here is done! thank you to everybody who made AIIDE a lovely experience for me, and to everyone who followed along my little live posting thread throughout the conference!

170 0

And now, for Best Student Paper, we have......... Generic Guard AI in Stealth Game with Composite Potential Fields! And for best paper, we have...... Speeding Up Narrative Planning Using Fog of War Pruning!! All awards are super well deserved :)

And now.... awards! We have 5 awards here, starting off with the Case Study award going to Mic Check! We also have best poster, going to Signals of Struggle! And best program committee member, going to Richard Zhao!

Location and Program Chair for 2026 are currently unknown, but will be announced very soon!

And now, closing off the conference, we have the closing remarks! This was the most attended in-person AIIDE since 2008, with 17* papers and 21 posters!

121 0

And with the end of the last paper session, we go to the StarCraft competition by Dave Churchill (photo missing, apologies) and the Workshop Report from Richard Zhao! @richardzhao.bsky.social Some interesting stats here, glad to see EXAG is thriving!

121 0

Closing off the session, we have my labmate and friend Dagmar Lofts @dagmarlofts.bsky.social presenting their DC on Process Centered Design for Creativity Support Tools. They're an amazing artist and researcher, and their approach to co-creativity is constantly inspiring. Great work!

230 0

And now we have a case study from Dave Goel, my labmate and friend! This work is incredibly cool: it's a method for extracting sprites and their underlying FSMs from a single gameplay video. You can see some of the learned code (using a DSL) in the last photo!

220 0

Next we have Yi-Chun Chen presenting on GameTileNet, a dataset of pixel art with semantic labels such as names and object types. This dataset seems like it would take quite a bit of work to construct, and could be useful for a number of tasks!

In the home stretch of the AIIDE marathon, we have our last paper session on Visuals, Audio, and More!! In the first talk I've ever barked at the screen during, we have Jonathan Morse presenting on Expressive Range Characterization for Text-to-Audio Models! Very fun talk!

230 0

Finally, we have Leonardo Villalobos-Arias on the Effect on Intrinsic Rewards on Agent Behaviours. I really appreciated the look at the behaviour patterns of the agents rather than just looking at the rewards. It was helpful to see what was going on!

Next we have Derin Gezgin on work that uses an LSTM network to play Sparrow Mahjong, a simplified Mahjong variant. I am curious about their evaluation here. I wonder how the network would perform against two PPO agents, as it's possible the worse agents help their agent win by "dealing in".

Starting off the posters of the session, we have Jay B. Nash remotely presenting on learning Space Invaders with Sparse Cosine Optimized Policy Evolution (SCOPE, good acronym) It's quite an interesting problem! I like systems that learn spaces, and I'm used to neural approaches doing this

Next, we have a DC from Manuel Eberhardinger on using program synthesis for various game AI problems such as agent explainability, control, and PCG. This work is very exciting to me, and I've had the pleasure of chatting with Manuel about program synthesis this week!

For the last full paper of the session, we have Kyota Kuboki presenting on Policies of Multiple Skill Levels for Better Strength Estimation in games like Go and Chess. I really appreciated the discussion of potential applications of the approach, such as teaching people how to play these games!

Next, we have Kaijie Xu presenting on stealth game AI using Composite Potential Fields. I've been following the stealth game work from this lab for a while, and love the visualization of the fields in the presentation!

Starting off our first paper session of the day on Agent Control, we have Sam Hill on the Declarative Speculation of Temporal Entities within a large-scale character simulation context. Sam is a clear presenter as always, even for those without a background in declarative programming!

I'm back with the last day of AIIDE 2025 live posting!

041 0

We're back to live (almost, because I was so engrossed I couldn't post during the talk) with Jonathan Schaeffer's keynote on the past, present, and potential future of Game AI. What an amazing retrospective with insight into the future! I'll be thinking about this one for a while.

Lastly we had Matthew McConnell presenting on a poster on generating levels for a Cosmic Express-like game using a genetic algorithm informed by player modelling! I got to try this at the demo session, and found the generated levels to be interesting and sometimes a bit challenging!

Following this DC talk, we had a double feature from Valentina Genoese-Zerbi, featuring her poster on adversarial story management(!), and her DC focusing on applications of this work to serious games. Very cool line of research!

Next, we had a DC talk from Mica Gardone on working toward a process of simulating narrative comprehension. I really enjoyed the speculative co-creative interface shown in the presentation (not pictured apologies)

Good morning, it's present Emily coming at you retroactively for the last paper session of AIIDE day 2! Up first, we had Honggu Kim presenting on CoDi, a prompt-based framework for generating stories that used multiple goal driven agents!

as an Official Emily Posting Update, I'll be your session chair for the last session of the day, so my posts may be a bit sporadic / late, but I'll get them all out there after I'm done my duties!

Finally, we have James Kwan presenting on using machine learning to predict combat outcomes from simulated data in the game Total War by Creative Assembly. This was a really cool look into how modelling work can be useful in an actual industry game, which I appreciated!

Next we have Kaijie Xu presenting an Adaptive Visual Navigation Assistant for 3D RPGs, a system that detects and highlights "spatial transition points" across souls-like games to help guide players at runtime and inform designers during development.

Next, we have Nabeeha Ali on a system for identifying Signs of Struggle in players playing a modded version of Terraria. This work is very exciting: even though it's just a case study for a single game, the fact that this is possible with just telemetry data is quite interesting. Very neat stuff!

@smshieldsy.bsky.social a tag that I missed for the last two, apologies!

Next we have Sam Shields (deja vu) presenting on visualizing the possibility spaces of DDA systems, melding together playtrace analysis and expressive range analysis! This seems like a really interesting tool, provided I'm not sure how close to existing in house tools this work is!

but I wonder about the notion of competitive integrity that might be lost here. Maybe there's a way to make the changes themself a mechanic (i.e. comeback mechanics like objective bounties in mobas?) (2/2)

We now have a Case Study by Sam Shields on FighterDDA, a testbed for automated balance that is set up to allow for a director agent to dynamically change the balance of a turn based RPG. It's interesting to consider that different players may want an entirely different game balance, (1/2)

Back from lunch, we have paper session #2 of the day: Player Modelling! We start off with Mica Gardone presenting on how to deal with incomplete and partially ordered information in the context of making sense of plans for agents.

Last but not least we have Kaylah Facey presenting on Game Behaviour Trees using Tile Re-write Rules, a new way of representing the behaviours of a game through these aforementioned rules. Very interesting case studies in this work, and I think the representation is quite neat!

Next up we have Mike Cook's @mtrc.bsky.social Pixie system which is a Unity plugin that can do mechanic generation in games that real people have actually made! Very cool, and related closely to the last DC in topic!

140 0

Next up we have a DC talk from my lab mate Johor Jara Gonzalez on Embedded Mechanics Generation! I think Johor's work is super cool, and if you're interested in it you should talk to him during the coffee break!

We have a DC talk by Daijin Yang on Responsible, Transparent, and Adaptive AI for Educational Game Design. I think this is a really interesting problem, and I really like the HCI angle Daijin took here!

Next up is a case study presented by Brooke Szajda on GameGoogle, a search engine that can be used to locate mechanics within games, such as searching for the concept of "basketball" within a game. Very cool stuff, I wonder about the potential for designing an association game using the system!