Automated workers still don't work.

[Gorbles]

Given that we're talking specifically about code, which is a pretty hot area of mine, I can confidently say that issues that arise due to implementation are probably the result of implementation, and not the person that originally thought up the mechanic and wrote it on a whiteboard.

The assumption you (or the other guy) are making, that the design deliberately sabotaged human Workers, is far less likely than the assumption that somebody coding something in a hundred-thousand line program made a mistake. I mean, we can point fingers all day. On what basis do you formulate your blame of design?

I mean heck, pointing fingers is completely irrelevant. We're talking about replicating past mechanics in a superior manner. If something was <not adequate> you can bet that that would be addressed in some way, either at the design level or at the code level. Are you suggesting that a subsequent implementation of a past mechanic would intentionally not be improved on?

 

[Browd]

Actually, I would surmise that the design for human worker automation always was intended to differ from AI worker automation -- in fact, had to differ in one fundamental respect. The AI worker automation code has the benefit of the AI strategic, economic and military goal subsystems, and the AI flavor weightings (RNG and all), all of which provide "guidance" to how the AI workers (and every other AI unit) should prioritize their activities in any given circumstance and turn, and therefore need to serve as "inputs" into the AI routines. The human worker automation system must, of necessity, use a one-size-fits-all approach, reflecting the developers' best guess at what a typical human player might want his or her workers to be doing at any given point in the game (I.e., in CiV, in the late game, devote every worker to upgrading all roads to railroads, because every player would want the RR production bonus in every city, right? Even in low-pop puppets that are only working trading posts?), subject to the few limitations we can specify in game options (such as not replacing improvements or removing terrain features). Of course, some code routines could serve double duty, for both purposes, but it should be expected that the AI and human worker automation will yield different worker "behaviors" (and not just due to flawed implementation of the human worker system, although that is certainly possible).

Anyway, you confirmed my supposition -- that you weren't making an assertion based on specific, inside knowledge about Firaxis design intent, but were making a (no doubt well-educated) supposition based on your software industry experience -- which is fair enough.

 

[Ar4i]

I've observed Automated workers to connect cities with road, so they actually do it. No idea when do they decide to do it though. In my last game it happened at some point in mid game.

 

[Gorbles]

Ah, sorry, I never intended to give the impression of inside knowledge. I don't have any and probably won't ever, for the record (Firaxis ranks as one of the developers I'd like to work at, but I don't plan on moving to the States, haha).

But actually your notion - that AI worker routines benefit from AI-visible factors that humans don't have access to (because they're inherent in our mental workings), is actually probably precisely the reason why human worker logic in the aforementioned game (Civ 4?) were more braindead than their AI counterparts.

They probably actually shared the same code, or close to it. Bereft of the variables that drive AI logic, they literally performed worse under the same conditions (when automated). If you could drive the variables into human worker logic they probably would've performed at a comparable level.

But that's problematic because that presumes you can just lift aspects of the AI logic and place it within the human class structure (in the code). For a game of Civilisation's complexity it probably simply wasn't doable, especially given the harsh timeframes software development often comes with (and especially games dev. - I'd never want to go through "crunch", personally, and my office benefits from a lack of reliance on overtime).

 

[Eji1700]

The ai needs workers to play...

And would be easier to handle if you had the ability for it to think in terms of specific areas (leashing) than it's current form where it tries to somehow make value judgements across civilizations.

It obviously needs to exist on some level, but frankly the AI needs a hell of a lot of things to work, and it's hardly doing a stellar job as is and that's not just because of poor worker decisions.

 

[mnf]

That seems...strange. If the AI is using the development team's best implementation of worker improvement logic (given city focus settings), why wouldn't you use the same code for automated player workers?

I think what I really wanted to say is that player worker automation should not contain any long-term decisions like what improvements go where. It should be clever enough to pose judgement (don't go near a furious bright red alien) but never decide for the player what to do.

Rather than "automated workers should...", I probably should have said "I want...".

While it's fine for AI play to let workers decide what to do, I'd like player worker automations to not involve that decision making. Players get to designate everything that the worker is supposed to do, with worker automation consisting of judgement calls like:

• I should first mine plot A then farm plot B because we're unhealthy already
• I should stop what I'm doing and farm that plot on the other side of the city because this alien can get me the next turn if I don't move

 

[TheMeInTeam]

Actually, I would surmise that the design for human worker automation always was intended to differ from AI worker automation

This. I doubt it was intentional sabotage, but the end result was that human workers, from 1.0 to Civ IV's final expansion release, would move adjacent to enemy borders and start improvements while automated, while AI workers would not do this. There's no viable reason that avoiding suicide should be specific to subsystems.

I can forgive them not making amazing improvement priorities, but I can't see justification for simply avoiding doing basic road pathing or creating improvements that align with city focus...and expending last movement next to things that can kill them is the kind of thing that doesn't belong in release versions of the game.

a game of Civilisation's complexity it probably simply wasn't doable

I'd buy this for tile optimization, but not worker suicides or refusal to construct some improvements outright (Civ IV did not have the latter though).

Players get to designate everything that the worker is supposed to do, with worker automation consisting of judgement calls like

That makes sense. It'd take extra work but it would be better. I considered the "not suicide, link cities, and make improvements based on focus" to be a low-effort, passable-but-not great solution. At least in that case, it's *less* of a trap mechanic, but what you're proposing would be more ideal.

 

[Gorbles]

I can think of any number of reasons why it happened, but obviously I completely agree that future implementations of such would aim to avoid replicating such an obvious bug.

As a very easy example that probably isn't the reason, let's assume AI workers have access to human border data, or parameters defining threat level. This is a part of the AI's calculations which the AI needs to be aware of every turn.

The human player doesn't need these variables. The human player can visibly observe borders by their graphical representation (which won't be closely-coupled to the data model, with any luck) and can work out if their Workers are heading towards a suicidal tile. Ergo the function which evaluates tiles to work will be passing in null (or 0, or whatever) for these values, while the AI step will pass in valid values.

This then results in erroneous behaviour!

That was my stance from the start - I can see how it happened in Civilisation IV, but I wouldn't see it happening if similar mechanics were adopted in the future, in a future game.

 

[SahintheFalcon]

It's nice to see that automated workers won't spam terrascapes or domes anymore.

The problem now is that they only seem to build farms and improve some resources while bizarrely leaving others untouched.

Now they will never EVER build roads or generators or nodes or domes or terrascapes or biowells for me, and it's been this way in over half a dozen games already.

How is this supposed to be a solution? It's as if the developers couldn't figure out how to fix them so they just stopped them from being able to build certain things altogether and called it a day.

Oh, and to people who keep saying "well you shouldn't be automating your workers anyway," if the feature is there, it should work.

Automated Workers:

1) They should never build roads. Roads cost maintenance. They are last priority in any build.
2) They should never build nodes. Nodes are useless, I would not build them anyway.
3) They should never build domes. Domes are useless.
4) They should never build terrascapes. Useless until very late game.
5) They should never build biowells. Biowells are only slightly useful in the mid game. Health isn't a problem in the late game.

I see no problem.

 

[Westwall]

So they shouldn't be fixed because the game is inundated with useless features?

Seems indicative of a wider problem that needs to be addressed.

Also, you haven't explained why generators are useless.