Pawn training: Part 1

Pawn Training and Inclinations

In this, I will hopefully provide some tips and information I've seen posted here throughout my stay in the general, and also share some of my own experiences with this game.

I feel like it's important to mention right off the bat here that most of what will be posted is just the above. I had the luck of finding an incredible community and relied solely on it for information when starting a new character and pawn after an original blind runthough of Dragon's Dogma near it's release. Mostly because of it I've played and kept coming back to this game for entirely too long now and have rented over 1,000 pawns, with accounts on either console. Pawns were one of the mechanics that really interested me in this game, and there are a lot of questions in the community about inclination settings and the seemingly accepted theory of pawn training.

In the off chance it should somehow be relevant, I'll mention a bit about my own personal playthrough. During my original game of Dragon's Dogma near it's release, I had a generic Mage pawn that I kind of hated and an Assassin Arisen that looked even worse than my current one does, and so when Dark Arisen released, I decided to just start over rather than importing my save. I began as a warrior with a warrior pawn, and decided on an all-melee playthrough, meaning I would never spend any time in a magic based vocation. I only maxed all vocations and played with my pawn in other vocations than Warrior well after level 200. I paid less attention at the time than I should have to stats and perfect augment setups. I maxed the vocations of fighter, warrior, ranger, and then Assassin on my Arisen, and spent the bulk of the original game outside of Everfall farming and then BBI playing as a warrior. My pawn spent enough levels in fighter and ranger to max them and then moved straight to warrior.

I also feel like it ties in to points I'll bring up later to mention that a majority of my time during my playthrough was spent without any pawn other than my own. Swapping between a dual Warrior and Assassin/Warrior team was and is a playstyle that I enjoy and utilized as much as possible before BBI stopped me in my tracks, and I stumbled upon the general and the community after shortly after DA's release.

Pawn training specifics were something I only heard after coming to the community, but it followed a pattern that I'd picked up on during my games- that the pawn learns specific habits from the Arisen and this particular knowledge affects it's performance in battle. I first heard it presented in the community that to better your pawn's performance, the Arisen should spend some time in the game using the pawns' desired skills and vocation. In my experience and in most instances, the pawns presented in the community, even when compared to similar pawns with matching inclination/beastiary knowledge on a specific enemy, preformed much more smoothly than any I'd hired thus far through searching the rift on my own.

But first, and perhaps just as importantly as the theory of training if not more so, a bit on inclinations. Inclinations are what is presented as the primary way that pawn AI dictates the pawn's actions in battle. The wiki presents them as follows:

Scather: Enter battles with enemies as they appear.
Medicant: Prioritize safety by healing your hero or removing their status ailments, using the 'Come' order during battle also helps.
Mitigator: Bring down weak targets first.
Challenger: Focus on long-range attackers and enemies with support abilities.
Utilitarian: Use support skills and enchantments.
Guardian: Use the 'Help' and 'Come' commands often, avoid using 'Go'.
Nexus: Quickly heal your pawns and remove their status ailments.
Pioneer: Use the 'Go' command when you're not in combat.
Acquisitor: Pick up items.
It also has few more battle specific actions presented by each inclination.

Throughout this page, there are various hints that an Arisen's playstyle and use of commands affect the pawn's inclination, even changing it, over time. So I should note that in the case of my main pawn before entering the community, this page and it's information is what I based my knowledge on. I had a vague sense of playstyle importance from my own experience. After experimenting with the more melee-based combinations, I found the results best suited to my playstyle were Scather/Utilitarian, and chose to keep the Warrior pawn at these inclinations as much as possible. Since my playstyle involved frequent usage of the commands and varying tactics against enemies, I carried inclination potions with me throughout the game and checked often to correct any changes.

Pawn training is a thing I personally believe holds some weight, and I'd like to try and shed some light on just how much. Later in the game and after becoming familiarized with BBI's monsters, I returned to my original playstyle outside of hires within the community and began trying out a few more unconventional ways of downing enemies.

I also began training my pawn in the way the community had recommended; with matching skills and vocations, although to be honest, a great deal of my time spent in the game prior had been as a warrior with 2 of his 3 skills. But beyond that, I was surprised on what pawns could apparently pick up on from their Arisen, along the way noticing the molding of and changes to my main pawn's combat behaviour. The pawn began to introduce new ways of dealing with enemies only after I had presented them first. More specifically, I would see the pawn preform a beneficial tactic or skill that I would use against an enemy, (a simple example being using Corona to knock an enemy down before using the Indomitable Lunge skill, or use of Exodus's i-frames)  and once the pawn was 'successful' with learning said action, it would attempt to repeat the action more often and to greater effect. It became a hobby and a way to keep the game fresh to try and impart new and as specific behaviour into my main pawn. I was lucky to receive extensive feedback over time about his performance, and in it recognized several of my own quirks.

To put it in the simplest terms possible- when I first started noticing the pawn imitating the behaviour of the Arisen, the thought was- well, can I maybe teach him to do those things I want him to do more, by preforming those actions often and to the best of my ability; and keep the pawn from undesirable behaviour not only by not using those skills or action as an Arisen, but by interfering with undesirable behaviour as much as I could. I had based the theory on 'success' of skills, and so whenever the pawn would start to preform an undesirable action, I would use opposing commands and if possible, prevent the Pawn's action from completion in it's entirety. And example I've used in the community is that I preferred the Warrior to not waste time grabbing enemies, an action he preformed often, so each time the pawn would do so, I would tell the pawn "Go!" while refusing to attack or kill the enemy. Since the original action no longer worked, the pawn would seem to consider the command, and then drop the enemy and attack it outright. In time, I noticed having to do this against the same enemies the pawn grabbed before less and less until in my game, the problem was near non-existent.

To follow up on, or see if skills I'd introduced had been implemented, I would switch from training the pawn as a warrior to the Assassin vocation and use the Invisibility skill to observe the pawn in action. To help the pawn recognize successful skill usage and modify the AI within it's inclination, I felt as though it would be best to let the pawn start and complete battles by themselves, with no help from the Arisen outside of the occasional commands like "Go!" and "Come!" for specific instances: for example, when you notice them targeted with an attack than can be avoided by moving themselves or when recognizing an exploitable enemy action, such as being stunlocked.

It may bear repeating that I feel like these results are best achieved by training with just the arisen and pawn, and that's how I did so, with no other pawns in your party, for reasons I will go into further detail on. It's also important, I feel, to re-mention that I kept the inclinations locked at Scather/Utilitarian myself regardless of my actions. I have heard theories regarding stuffing a certain number of potions at once can "lock" inclinations into place and keep them from budging, as well as using them to set a hidden tertiary inclination. Here's my take on that:

I never did it and thus have no idea.

I am unable to find a source or locate the poster on the following info in quotations, so please take it as you will. However, this was posted in the community often a while ago by one of our more science-minded members, and was supposedly the recollection of a conversation held at a DD promo event with an AI developer. Since it coincides with my own personal experience I've decided to post it here, but since I cannot provide a source, I will leave it to your discretion whether or not to regard it as accurate.


"This is from an Anon some weeks ago, said he went to a DD promo event and chatted with the lead AI dev. I asked him to post what he remembered and this is what I got:

"As far as AI goes, the general gist was what I already described in a prior post. The way the flags work in particular is as an over-ride. They did this because in play testing, people who didn't manage their pawns much created completely ineffective, useless pawns, who learned all the wrong things. So they added the star system both as a way to ensure pawns always attain some level of effectiveness, while also giving an easy way to compare pawn knowledge while in the rift, something they had a tough time finding a solution with before. An example he used was something like this: when a pawn sees a wolf, they make a decision about how to attack it, they consider multiple factors, such as distance, and their class and abilities into account. They will use whatever they most learn is effective based on experience. However, when they added the flags in, for example, once they hit a wolf with fire, fire becomes the default assumption for that enemy type. So, now, 'wolves hate fire'. Whenever facing a wolf from now on, they will heavily favor and lean on fire techniques more than any other. This sort of flag overrides their core AI. However, this is only by setting 'fire' as the most effective thing by default, without them having to learn it. So say for example, you remove that pawns fire skills, and give them a skill that has much higher DPS and effectiveness against wolves, despite not being fire, and they use it enough that they learn how effective it is against wolves, even if you switch them back with a mix of fire skills, they will now consider and judge this new skill based on its effectiveness, IF that skill matches or exceeds the effectiveness of what it witnessed with fire. All 'i dont know' solutions by the pawn will default to fire in the particular case of that enemy."

"Also, whenever a pawn says 'I learned a new technique against that foe', it means they learned something that has nothing to do with flags or triggers- all the flags or triggers have specific quotes attached to it, such as 'its kind hates Ice and fire both!'. "I learned a new technique against that foe" is the blanket term for anything a pawn could learn that the developers either didn't account for, or didnt associate with a specific trigger. In addition to this, Inclinations have an effect on behavior as a 'modifier', they didnt go into much more detail on that, other than it modifies actions or how a pawn does actions after it already decides to do them based off of the AI or the triggers.

Pawns also mimic their masters in attempts to learn new behaviors. Outside of attempting things the master does, They wont try new things unless given an order, in which guess they try to guess or evaluate your desire based on the order. If it is successful, they do it more often, even without orders. If it fails, they do it less often, even with orders.

However, they will try new things in a 'reactionary' sense- they wont try new things for no reason, but if encountered with a problem with a challenge, and they cant decide on an answer, they will try whatever comes to mind, and then evaluate the results afterwords. Of course, this doesn't work if their mistake causes you to die, because they never get to consider how they failed if you reload a save file, which is something they are looking to improve in the future they said.

On multiplayer, someone who talked with Itsuno told me that Itsuno was considering multiplayer, but only in a limited fashion, from the sounds of things, he may be considering the Everfall to be a multiplayer dungeon next time around. But this is heresay, I never got the opportunity to talk to Itsuno myself."

So that's that.


As far as to why I feel it's so important to take your pawn out alone when training and to let them fight battles on their own. "Do pawns learn from/change their behaviour around other pawns and Arisen?"

Short answer: Yes, they change their behaviour around other pawns.

Long answer:
In this particular experiment I have hired four sorcerer pawns and four ranger pawns, for no other reason than they are (supposed to be) ranged attackers and this makes it easier to observe their behaviour while sitting on the sidelines as an invisible assassin. Since this control focuses on group AI affecting individual pawn behavior rather than inclination testing, each Ranger had matching inclinations, Challenger/Scather, and each Sorcerer was Utilitarian/Scather. Each shared three stars in the beastiary against the particular enemies I took them to fight and observe- the direwolves, snow harpies, and chimeras. I took them each out individually, lacking a fourth pawn, as well as in groups of two rangers, two sorcerers, a ranger and a sorcerer, etc. None of these pawns are in the community, so it may be unfair without having the owner's permission to post their tags here.

To try and make this short, when solo ran, the ranger pawns freely used both their bows and daggers, although unsurprisingly to me, some preformed these skills much more smoothly than others. One in particular favored daggers, hardly ever using their bow even when trying to target the harpies, and opted to cancel their attacks often. The other pawns ranger pawns would cancel to a similar extent, but not as pronounced. This is something I've seen a lot with Ranger and sorcerer pawns- they like their space and will spend a lot of time dodging and trying to "line up" a particular attack with an enemy. It often looks like they may not be attacking because the pawn is trying to find a better approach. This coincides with earlier theory that pawn AI considers not only inclination, but their successful skills, weapon reach/ability, and prior knowledge of successful skills and behavior before selecting their course of action.

However, when observing the ranger pawns (and sorcerer pawns) in groups of two, they would fall into sync with the other pawns in the party and react to situations differently than they had alone. They'd imitate the action of the other pawn. When the ranger that was proficient with their bow started to fire arrows, the other pawn that prior showed lack of it's use began to as well. The sorcerers with Uti would notice other pawns and attack and cancel with spells accordingly. When ranger/sorcerer, the two pawns would stay close together, determining a safe distance and attacking similarly and in sync; such as the rangers seemingly not using bow skills while a large 'world ending" spell was cast. The rangers all climbed the chimera in a similar fashion as well when one did, when only Ranger #3 climbed the enemy during their solo run. Hopefully in the future I will be able to provide additional recording of instances like this.

Another hint that pawns and party composition affect your pawn's AI and combat behavior is verbal hints related to techniques against foes. I have my Warrior set to be pretty talkative- which probably pisses a lot of people who hire him off, but help me out to determine my pawn's 'train of thought' so to speak. Even so, outside of random and 'landmark/checkpoint' pawn babble such as rocks and trees, the pawn does not speak without clear motive to do so. Their vocalizations and more importantly the timing of such, such as initial recognition of/targeting an enemy, helps me to determine their next course of action and their trigger phrases for learning new behaviour.  Pawns express themselves differently around different party compositions.
An example is, while taking out a party that consisted of myself, a Ranger at the time, as well as a hired fighter and sorcerer, my warrior helpfully reminded the party to "Block the incoming attacks" from the approaching Garm with a shield. This is a tactic I have never used. I had not faced Garms as a fighter nor did I use shields to block ever in my assassin playstyle. The only conclusion I could draw, (since the main Warrior pawn was clearly addressing the other fighter in the party and not the ranger Arisen or the Sorcerer,) is that the pawn learned this behaviour and how to address it in this way while in someone else's party, furthering his foe knowledge and AI triggers. Without a fighter in the party, I never heard this line.

As far as another Arisen specifically influencing your pawn's behaviour when hiring your pawn, it was also theorized within the community a while back that hires and other Arisen using commands while your pawn was in their party did in fact change pawns' behaviour and inclination. Players who had frequent hires started to notice their pawns returning from batches of hires and behaving poorly/switching inclinations immediately thereafter. The next pastebin will focus on some of my findings on this. What I've done is taken a control group of warriors from the community with levels varying in brackets between 1-50, 50-100, 150-200, and 200, both "trained" and "untrained", and with the owner's permission I will, by brining my main pawn, interact with these community pawns and train using the exact same techniques and methods I used on my own warrior main pawn, and attempting to narrow down/pinpoint specific effects my actions have on their pawns over time, if there are any.

Finally, I will be taking pawns who have been "trained" and recording their actions alongside random rift pawns with matching inclinations and foe knowledge to hopefully show individualization and effectiveness that training brings to pawns. I will also attempt to show details of playstyle training and command use as described here.

None of this would have been possible without the help of an outstanding community, and I want to express my thanks to them for helping me enjoy this game quite as long as I have. All of this, therefore, is presented not as gospel or as unchangeable fact, but rather personal experience and experiments to help share with that community and the questions they have regarding pawn training. It's all in the spirit of jolly co-operation and I'd love the chance to learn new or contradictory info outside of what I've presented here.

Thank you for your support and hires.