[1.0.12.43] AI does not want to select techs with high score

[Infixo]

Here is the situation. The Future Era is like this: <seasteads> => all others techs => <offworld mission>. So, ex. those 2 techs, all others have the same cost. In turn 301 Seasteads are finished and AI needs to make a decision on what to research next.
Here is the log:

301, 0, Tech, TECH_SEASTEADS, 0.0, OWNED, -1
301, 0, Tech, TECH_ADVANCED_AI, 155.3, , 4
301, 0, Tech, TECH_ADVANCED_POWER_CELLS, 155.3, , 4
301, 0, Tech, TECH_CYBERNETICS, 155.3, , 4
301, 0, Tech, TECH_SMART_MATERIALS, 11875.2, , 4
301, 0, Tech, TECH_PREDICTIVE_SYSTEMS, 155.3, , 4
301, 0, Tech, TECH_OFFWORLD_MISSION, 29678.0, GOAL, 23
301, 0, Tech, TECH_FUTURE_TECH, 19842.5, , 28

This Civ is following Science Victory strategy, and goal set as Offworld Mission is ok. Also - Smart Materials are scored high, because there's Exoplanet Expedition.
Ok, next turn:

302, 0, Tech, TECH_ADVANCED_AI, 155.3, RESEARCHING, 4
302, 0, Tech, TECH_ADVANCED_POWER_CELLS, 146.1, , 6
302, 0, Tech, TECH_CYBERNETICS, 146.1, , 6
302, 0, Tech, TECH_SMART_MATERIALS, 11171.4, , 6
302, 0, Tech, TECH_PREDICTIVE_SYSTEMS, 146.1, , 6
302, 0, Tech, TECH_OFFWORLD_MISSION, 23634.6, GOAL, 33
302, 0, Tech, TECH_FUTURE_TECH, 15787.8, , 40

AI chose Advanced AI with score 155 instead of Smart Materials with score 11170

After that it chose Cybernetics.

308, 0, Tech, TECH_ADVANCED_AI, 0.0, OWNED, -1
308, 0, Tech, TECH_ADVANCED_POWER_CELLS, 134.2, , 8
308, 0, Tech, TECH_CYBERNETICS, 146.1, RESEARCHING, 6
308, 0, Tech, TECH_SMART_MATERIALS, 10268.1, , 8
308, 0, Tech, TECH_PREDICTIVE_SYSTEMS, 134.2, , 8
308, 0, Tech, TECH_OFFWORLD_MISSION, 20811.1, GOAL, 36
308, 0, Tech, TECH_FUTURE_TECH, 13901.3, , 45

Then Advanced Power Cells.

314, 0, Tech, TECH_ADVANCED_AI, 0.0, OWNED, -1
314, 0, Tech, TECH_ADVANCED_POWER_CELLS, 139.8, RESEARCHING, 7
314, 0, Tech, TECH_CYBERNETICS, 0.0, OWNED, -1
314, 0, Tech, TECH_SMART_MATERIALS, 10693.1, , 7
314, 0, Tech, TECH_PREDICTIVE_SYSTEMS, 139.8, , 7
314, 0, Tech, TECH_OFFWORLD_MISSION, 26621.6, GOAL, 26
314, 0, Tech, TECH_FUTURE_TECH, 17793.2, , 34

After that Smart Materials finally.

I have checked another Civ that was leading in science and it followed the same order despite scoring Smart Materials way higher.

I suppose this might be the reason why AI is unable to effectively (i.e. quickly) get the Science Victory.
Once the tech is researched, it does the projects, even the laser ones. But it just can't get to it in time.

I started investigating this when in another game Japan launched the Exoplanet and lost to China that launched the mission 15 turns after Japan. What happened? China had Offworld Mission tech and built 3 lasers, effectively getting the victory. I was wondering why Japan didn't build them? Well, because they didn't have the tech. And instead of researching it - went for some other techs, not relevant to the science victory.

 

[pokiehl]

Interesting find.

To make sure I understand you properly: the technologies are "scored" appropriately for the AI's strategy, but the bug is that the AI is ignoring this and going for techs that are scored lower?

 

[Infixo]

This is a rare situation with multiple techs that have the same cost. Usually AI goes for the highest score. I have no theory yet why the above happens. It just doesn’t make sense.

 

[MrRadar]

What does make even less sense is that AI will almost always get Smart Materials and might even get Future tech a few times over before it researches Nanotechnology. Unless Nanotech by the mercy of RNG end tree is made a direct prerequisite for SM, AI invariably ignores the whole Plastics-Nanotech line until very late. They may be on the Moon by t240 on Deity, but then they just research everything except this line, resulting in 40-50 turn pause in the space race. Maybe already Plastics is scored too low? This is the first modern tech AI shows real aversion to.
And they will always build Lagrange lasers only, if they have aluminium. Never ever I've seen them building a terrestrial laser without tinkering with the code. Both laser projects are not priority, btw. as far as I remember from observing AI-only games during the excessive science bias times.