The previous two posts in this series described how research was done to find if language clues could predict when a betrayal was about to happen, and what the researchers found. One thing they found was that humans are bad at spotting impending betrayal in a game of Diplomacy.

There are clear clues. A betrayer is likely to be more polite than a victim, although a victim might suddenly become much more polite before a stab. A victim tends to use more planning language than the betrayer, while a betrayer is likely to use more persuasive language. A betrayer tends to be more communicative, although if they are using complex arguments and stating opinions based on evidence it is not likely to happen immediately.

What can we take from this study? Can we use it to become more aware of when we are facing the threat of a stab? And can we use it to help to mask the stab more effectively?

Armoured Against a Stab

The thing is, if we’re going to use these linguistic clues, we need to take much more notice of a player’s writing style than I believe we usually do. Given that a lot of us are bad at reading the clues from the board, this is a tough ask!

What we can do is:

1. Be more polite. The research showed that victims were less polite in their communications than the betrayer. This could well be just something that happens naturally – players tend to be more prepared to stab a less-polite person. However, suddenly increasing the level of politeness seems to indicate that you’re more likely to be stabbed! So be consistently more polite. Don’t demand actions, ask about them.
2. Keep planning discussions to a minimum. I find this hard. If I have a plan of action, I want to let my ally know exactly what it is. And, often, I don’t have too much of a problem with doing so. This is perhaps because the planning imbalance is more indicative of the fact that the betrayer is not invested in the alliance any more. However, perhaps a better approach is to discuss what you want to happen simply and, if the other player engages with that, then you can expand.
3. Watch for changes in complexity. This isn’t just in planning. If a player begins to write less involved messages it is an indication that they are less interested in the alliance.
4. Watch out for persuasion rather than complexity. If your ally begins to work on persuading you that they’re completely invested in the alliance, that they have no need to stab you, etc then this indicates that they are more likely to be planning a stab. This isn’t about the amount they write, but what they write about.

Hiding the Stab

Perhaps it isn’t quite as necessary to take notice of this part, given that other players are more than likely not going to see the stab coming, but there are certainly lessons to be learned. After all, there are times when the board is a pretty clear indicator that the stab is coming. So, what can you do to try to make sure your victim doesn’t suspect the stab in the first place?

1. Maintain the complexity in your messages. It’s easy, when you’re planning to stab someone, to simply stop thinking about the alliance and its aims. You’ve other things taking up your time, after all. But don’t. It isn’t about keeping messages long, but about considering what the other player wants to see in your replies. Answer their questions, respond to their requests, add a note or two of your own. Don’t replace complexity with raw persuasion… that’s a give away. Simply being involved in the discussion is persuasion enough.
2. Keep your persuasion to a minimum. When we’re about to stab someone, we want them to believe it isn’t going to happen, so we tend to become more persuasive. After all, if you can convince your victim that you’re definitely not interested in stabbing them, job done, right? Well, an increase in persuasiveness is an indicator that you’re more likely to stab. Not that it doesn’t work – apparently it does! But better to show that you’re invested in the alliance with substance rather than promises.
3. Control your positivity. This is similar to the above. Sometimes we over-sell the advantages of maintaining the alliance rather than providing promises. After all, it’s all persuasion: the latter is active persuasion, the former is an attempt at subliminal persuasion, persuasion by stressing how successful you’re going to be. Again, much better to be invested in the conversation than replacing substance with noise.

Consistency Succeeds

If you’re looking to hide a stab in messages, the trick is to remain consistent. When you were building the alliance, you were all about how you could help each other and what moves to make to achieve your common goals. You didn’t need to be persuasive because you had things in common.

Now, when you’re considering ending the alliance, your goals have changed. You can’t allow that to come through in your messages. The other player is, hopefully, still invested in the alliance, they still have things to achieve that you can help with and that should help you. You have to make them see that you’re still as invested as you ever were and changing the level of discussion, changing the type of discussion, should set off alarm bells.

On the other side of the board, if your ally changes something in the way they’re messaging you, you should be hearing that alarm. What’s changed? Why has it changed? Have they upped the intensity of their efforts to persuade you of something, rather than discussing the meat of the problem? Uh-oh!

Always keep in mind what the other player wants. You can avoid a stab by working towards that with them; and you can increase the chances of success when stabbing by seeming to be invested in the other player’s objectives. If the other person doesn’t seem to be interested in pushing you forward with them, then find your armour – you’re going to need it.

POSTS IN THIS SERIES:

• Harbingers of Betrayal: Introduction
• Harbingers of Betrayal: Findings
• Harbingers of Betrayal: Lessons

In the previous post in this series I explained that, in July 2015, a paper was published that had looked at whether there were any clues in language that might indicate a friendship was about to end. This research was based on a study of online Diplomacy games.

So, what did they find out?

Well, first, how did they look for clues? “To ensure that we are studying conversational patterns that occur only when the two individuals in the dyad [2-player pairing] are ostensibly being friends, we only extract features from the messages exchanged before the last act of friendship. … we can only hope for subtle linguistic cues: if there were salient linguistic signals, then the victim would notice and preempt the betrayal.”

So they’re looking at the messages sent before the act of betrayal, the stab. They compare these with other conversations between 2 players at a similar stage in the game, from a similar length of alliance, but where the alliance continued.

By comparing the patterns in language from these two types of game, they came up with the following evidence.

Sincerely Insincere

Positivity Imbalance

It isn’t surprising that alliances are built on positive interactions between players. You’re not going to willingly ally with someone who tells you how bad things have gone, are going or will go. Alliances mimic friendships; it’s hard to like a moper.

What they found was that a player who was about to betray another tended to use a lot of positive sentiment. In fact (and this is surprising) they would use more positive sentiment than players who were in an alliance that lasted. Why should this be? As the authors suggest, possibly over-compensation. If a player is about to betray you, they could be using excessive positivity to mask the impending stab.

Planning and Persuasion

They looked at the occurrence of language of persuasion & discussion, and at language indicating planning activity. What they found was that a victim was likely to have more planning related language in messages than the betrayer. This may be because the betrayer doesn’t see any future in the relationship and has subconsciously given up.

On the other hand, what they found was that a betrayer would use more persuasive language than a victim. The betrayer may have stopped thinking in terms of planning, but they’re certainly trying to persuade the victim of something… how strong the alliance is, perhaps?

Politeness

It’s not difficult to see politeness in messages. Every message should be polite. However, again, they found something which seems less predictable: that before an alliance broke down the victim used fewer politeness markers than the betrayer. Maybe that’s because the betrayer, knowing the power they have, is more polite as a way of distracting from the coming stab or possibly is softening the victim up. Of course, it could be that the betrayer is preparing to stab because they’ve got fed up of the other players impoliteness!

Patter

They looked at the number of messages sent, the number of sentences per message, and the number of words per sentence. What they found was that there was an imbalance between the talkativeness of the two players when a betrayal is about to take place. What they saw was that a betrayer is likely to use more sentences per message than a victim.

Warning Signs

The researchers built a model, using their findings, to see if these findings would allow someone to predict that a stab was coming. They presented this to players and ran a classification model.

With players, the clues were missed. People were unable to see the betrayal coming. In fact, the results suggest that you may as well toss a coin to decide whether someone is likely to betray you or not. We all know this: it’s where the paranoia in Diplomacy comes from.

However, the classifier was able to predict betrayal at a level significantly higher than chance. This tells us that, given a raised level of awareness, we should be able to better predict when someone is going to betray us.

What are the clues? “Overall, the selected linguistic features capture a consistent signal that characterises people’s language when they are about to betray: they tend to plan less than their victims, use less structure in their communication, and are overly positive.” We can also see that they will be increasingly persuasive and write longer messages.

Of course, some of this may just be the way people communicate. Perhaps it’s even more to do with the profile of a player who is more likely to stab and one who is less likely to stab.

So we need more information and the researchers looked back over previous turns. What they found was that betrayers begin to show a change up to two turns before the act of betrayal. This isn’t surprising: the best stabs are planned ahead of time.

Positivity imbalances occur just preceding the stab. If someone is still expressing opinions through ‘claims’ (that is, stating what you think through perceived evidence), they’re not likely to stab you immediately. Betrayers become increasingly more talkative than their victim the closer the stab looms.

They also found that, if one player tends to talk more about plans, they are more likely to be stabbed, possibly because of the pressure of planning having a negative effect on the betrayer. Interestingly, if a player suddenly becomes more polite, then they’re likely to be betrayed!

So what can we learn from all this? Well, first, we’re bad at predicting when a stab will happen. This isn’t news. Even leaving aside our inability to read the clues the team looked at, some of us are bad at reading the board and predicting when a stab is likely to happen. Perhaps we prefer to snuggle beneath the comfort blanket of a long-term, on-going alliance than to peek out head out to see if we have to throw the blanket off.

POSTS IN THIS SERIES:

• Harbingers of Betrayal: Introduction
• Harbingers of Betrayal: Findings
• Harbingers of Betrayal: Lessons

In July 2015 the Association for Computational Linguistics (ACL) published a research paper called “Linguistic Harbingers of Betrayal: A Case Study on an Online Strategy Game” and, surprise surprise, that game was Diplomacy.

It makes for interesting reading, although Dippyists with some experience would probably say that the findings are fairly intuitive. And, let’s face it, any research paper that manages to get the word ‘harbinger’ into a title is worth a read anyway!

I thought, though, that I’d write about the paper and its findings. It would probably make sense for you to read the paper first!

I wanted to start with an introduction to what the paper was researching. It’s probably pretty clear from the title that it’s looking at what clues there are in a person’s language that indicate betrayal.

If this paper is accurate, it can be of great use to you as a Dip player. Not only would it help to anticipate an impending stab, it could also be used to mask one! Imagine that! Not only are you planning to stab your erstwhile ally but you’re going to successfully hide that fact from them! Invulnerability, here I come!

Well, unless of course your opponent has read the paper, too, and they’re applying the lessons from it, in which case you won’t see their stab coming. Dammit.

Definitions

Let’s start by looking at some definitions for words and terms that come up in the paper.

• “… we explore linguistic cues that presage such transitions by studying dyadic interactions.” Dyadic simply means ‘two persons” so they’re looking at messages between two people. Similarly, the word dyad is used throughout, by which they mean a 2-player conversation.
• Stable friendships: These are when two players have shown evidence of “at least two consecutive and reciprocated acts of friendships that span at least three seasons in game time. We also check that no more than five seasons pass between two acts of friendships, as friendships can fade.” This is based on on-the-board actions. Of course, two players could be allied without any on the board action at all but this is difficult to evidence accurately.
• Betrayals: “Betrayals are established and reciprocal friendships that end with at least two hostile acts. The person initiating the first of these hostile acts is the betrayer, while the other person is the victim.”

These two definitions are important for the research they’re doing, which is aimed at comparing the language used by and messages exchanged between two ‘friends’ and those between two friends who become ‘enemies’.

Their description of Diplomacy

If they’re using Diplomacy as the basis of their research, it follows that they must explain the game. Unfortunately, there are some errors in how they describe it that a picky Dippyist – me – could find a little irritating.

First, they don’t accurately describe the objective of the game: “The goal of the game … is to capture all of the territories on the game board.” Of course, it isn’t: the goal is to control the majority of supply centres on the board.

In case you’re new here, and you’re unsure of the difference between the words ‘capture’ and ‘control’ here it is. You can capture an SC by moving into the space the SC occupies. However, if you do this in a Spring turn, you won’t own or control that SC; you are simply occupying the space. Your unit needs to remain in occupation until after the Fall Retreats phase. Only then do you own or control the SC.

Second, they don’t accurately describe how the game is structured: “Each season consists of two alternating phases: diplomacy—the players communicate to form strategies—and orders—the players submit their moves for the season.”

In traditional rules, the game is split into years (1901, 1902, etc), turns (or seasons if you like) (Spring and Fall), and phases (Moves, Retreats and, in the Fall turn, Adjustments). The modern rules go further, dividing the phases into Diplomacy, Orders, Resolution (I’d prefer ‘Adjudication’ but meh), Retreats and, in Fall, Adjustments. What they’ve done is split the main phase – Moves – into three. This makes sense although it’s quite likely that players will write their orders throughout the Diplomacy phase.

It’s a small error and an understandable one – the authors aren’t interested in the Resolution, Retreats or Adjustment phases because there’s no communication in these phases under the game rules (although there is likely to be when played online).

Third, they over-simplify some of the explanations about the game: “Instead of moving, a unit can
support another unit; large armies can be created through intricate networks of support. The side
with the largest army wins the battle.”

This is being very picky, I know, but it ignores the action of fleets. The majority of SCs on the board are accessible to fleets: there are 34 SCs and 27 of them are ‘coastal’ SCs. Oh, well.

They seem to have their timings wrong: “We use games from two popular online platforms for playing Diplomacy. The average season of an online Diplomacy game lasts nine days.” This was before the current preference for shorter deadlines but, even so, turns online rarely last that long… unless they were concentrating on games with long enough deadlines to allow more complex communication.

These differences don’t affect the outcome of the research, though, and I only mention them because, well, they’re acknowledged inaccuracies among Dip players. Some of this is about simplifying the terminology or explanation of the rules so that they don’t hamper the reason for the article in the first place.

So, let’s move on to the findings.

POSTS IN THIS SERIES:

• Harbingers of Betrayal: Introduction
• Harbingers of Betrayal: Findings
• Harbingers of Betrayal: Lessons