What makes a good Diplomacy lie and can you pick one when you hear it? The guys catchup with academics Jordan Boyd-Graber and Denis Peskov about their paper on lying and Diplomacy. Plus Sloth, a player in the study, also joins us. And don't forget Christmas 2020 and much more.
Intro
The guys have got the recorder working again, although they're trying something different so it doesn't cut out again. They discuss their drinks and venue (0 mins 10 secs)
They throw around a couple of jokes (2 mins 20 secs)
The guys set the scene for the interview today with Jordan Boyd-Graber, Denis Peskov and Sloth about their paper, It Takes Two to Lie: One to Lie and One to Listen. The paper discusses machine learning and the ability to detect lies through Diplomacy (3 mins 10 secs)

Interview with with Jordan Boyd-Graber, Denis Peskov and Sloth
NOTE: Amby tried recording this interview in stereo for something different. Not sure it really works. Gives us your feedback (either on this page's comments area or by contacting us)
Kaner and Amby welcome Jordan Boyd-Graber, Denis Peskov and Sloth to the show before Amby asks Jordan to give the background to the paper and why it concentrated on Diplomacy as a means of concerted lying (4 mins 40 secs)
Denis discusses the process around using identifying messages and whether the sender lied or whether the receiver thought it was a lie. He also touches on his discovery and now love of the game (8 mins)
Amby tries bringing Sloth's perspectives into the game as a player in the study and his thoughts on lie detection (11 mins 30 secs)
Amby asks Denis about player prediction of lies vs the artificial intelligence's ability. Jordan gives his perspective (14 mins 15 secs)
Amby goes onto ask about factors to indicate more likelihood to lie or tell the truth (17 mins 15 secs)
They explore whether there were points of time in the game when lying was more likely to occur (27 mins 30 secs)
Amby reflects on the difference in physical presence to the game in face-to-face vs online, he asks about people blaming real life as a crutch for a lie. They also discuss people's perceptions of what is a lie. Sloth points out how people don't mind lying, they just don't want to get caught lying (30 mins 30 secs)
Kaner asks about deceitful language online and its impact on machine language into the future (34 mins 20 secs)
Kaner goes onto ask about whether anyone has approached some of the big Diplomacy websites about their datasets (40 mins 20 secs)
Amby gives his thoughts about the methodology using both a combination of the words people use (eg a lie) and the ability to stab (42 mins 40 secs)
Jordan needs to leave the interview. Amby asks if there were any zones people were more likely to lie about. Denis asks his own questions to Kaner and Amby on who gets stabbed the most, and also which country was more likely to stab (45 mins 50 secs)
Denis discusses some of the lies sent through from Kaner and Amby in their recent games to see if the IA could predict if the lies were lies. You can view the questions and predictions here (49 mins 40 secs)
Sloth asks about elements of truth. Denis discusses the data rather than the emotional aspects (55 mins 30 secs)
Amby asks about future research plans in this space (57 mins 20 secs)
Amby starts wrapping up the interview. Just a reminder, you can read the paper if you want to learn more (58 mins 50 secs)
The guys return back and discuss what they got out of the interview (59 mins 40 secs)
Kaner flags how good it would be if webDip and other sites had an option to be part of the lying research so players keep records on whether they lie/predict a lie. Amby discusses the possibility of league tables for Best Diplomacy Liar or Best Predictor of Lies (1 hr 7 mins)
Amby asks about how could you apply the same Lying methodology beyond text based chat, extending to voice communication games eg on Discord, and whether their lying styles change (1 hr 11 mins)

It's Christmas time
The guys keep t