Nei fenomeni sociali i costituenti di base sono gli esseri umani, cioè individui complessi che interagiscono e cooperano con un numero limitato di pari rispetto al numero totale di persone nel sistema, con lo scopo di raggiungere un obiettivo comune. Comprendere e modellare i fenomeni collaborativi è importante per spiegare e prevedere ciò che accade realmente nella società in cui viviamo, in termini di cultura, leggi, contesti sociali e molto altro. Più recentemente, abbiamo anche iniziato a interagire con gli altri in un ambiente online, ad esempio possiamo condividere i nostri aggiornamenti personali sui siti di social networking, impegnarci e commentare i post dei blog, ma anche partecipare a comunità virtuali, come il caso di "Reddit Place", un esperimento sociale creato da uno dei più famosi social network, Reddit, nel 2017. I partecipanti (utenti) avevano accesso a una tela da disegno dove potevano cam- biare il colore di un pixel a ogni intervallo di tempo prefissato. Gli utenti non erano raggruppati in squadre e non avevano obiettivi specifici, tuttavia si sono organizzati in una struttura sociale coesa e hanno collaborato alla creazione di una moltitudine di artwork. Il vantaggio delle interazioni sociali online è che possono essere facilmente registrate sotto forma di dataset e poi studiate per scopi scientifici a beneficio della ricerca; infatti, in questo lavoro abbiamo effettuato uno studio su un set di dati comprendente più di 16M di azioni degli utenti, registrate nell’esperimento sociale online, con l’obiettivo di scoprire come le persone collaborano e si organizzano nei giochi sociali online. Poiché i modelli di collaborazione sono difficili da catturare quando le relazioni tra gli attori non sono direttamente osservabili, in questa tesi utilizzeremo strumenti di teoria dell’informazione per identificare specifici modelli di collaborazione; uno di questi è la Partial Entropy Decomposition (PED) che ci permette di distinguere il comportamento degli utenti che collaborano da quelli che non lo fanno. Per comprendere le dinamiche della comunità, abbiamo realizzato diverse definizioni di interazione tra gli utenti, alcune basate solo sul posizionamento dei loro pixel altre che includevano anche il loro colore. Inoltre, grazie al timelapse dell’esperimento, abbiamo notato come i diversi artwork avessero comportamenti diversi durante tutto il tempo: alcuni sparivano, mentre altri rimanevano nelle stesse posizioni, indicando probabilmente la presenza di un conflitto, nel primo caso, e di relazioni pacifiche, nel secondo. Infine, abbiamo cercato di replicare la dinamica della tela su cui gli utenti inter- agiscono generando serie temporali sintetiche utilizzando l’algoritmo di Gillespie e le nostre definizioni di interazioni, confrontando i vari risultati con le misure di entropia citate in precedenza e dimostrando come la PED sia uno strumento estremamente utile per identificare i pattern di collaborazione anche in un contesto online. Questo lavoro mostra quanto possa essere complesso il dialogo tra la struttura di un sistema e il suo comportamento cooperativo emergente, ma offre anche alcuni risultati interessanti, soprattutto per quanto riguarda l’uso della PED in un ambiente online di interazioni high-order, come r\Place.
In social phenomena the basic constituents are humans, i.e., complex individuals who interact and cooperate with a limited number of peers compared to the total number of people in the system, with the aim of achieving a common goal. Understanding and modeling collaborative phenomena is important to explain and predict what really happens in the society we live in, in terms of culture, laws, social contexts and much more. More recently, we have also begun to interact with others in an online setting, for example we can share our personal updates on social networking sites, engage with and comment on blog entries, but also participate in virtual communities, such as the case of "Reddit Place", a social experiment created by one of the most famous social networks, Reddit, in 2017. Participants (users) had access to a drawing canvas where they could change the color of one pixel at every fixed time interval. Users were not grouped in teams nor were given any specific goals, yet they organized themselves into a cohesive social structure and collaborated to the creation of a multitude of artworks. The advantage of online social interactions is that they can be easily recorded in the form of dataset and then studied for scientific purposes benefiting research, in fact, in this work we performed a study on a dataset comprising more than 16M user actions, recorded on the online social experiment, with the goal of discovering how people collaborate and organize on online social games. Since collaboration patterns are difficult to capture when the relationships between actors are not directly observable, in this thesis we are going to use information theory tools in order to identify specific patterns of collaboration; one of these is Partial Entropy Decomposition (PED) which allows us to distinguish the behavior of users who cooperate from those who do not. To understand the community dynamics, we, made several definitions of interac- tion between users, some based only on the placement of their pixels others that also included their color. Moreover, thanks to the timelapse of the experiment, we noticed how different artworks had different behaviours during the whole time: some of these were disappearing, while others remained in the same positions, probably indicating the presence of a conflict, in the former case, and peaceful relations, in the latter. Finally, we attempted to replicate the dynamics of the canvas on which users inter- act by generating synthetic timeseries using Gillespie’s algorithm and our definitions of interactions, comparing the various results with the entropy measures mentioned earlier showing how PED is an extremely useful tool to identify collaboration patterns also in an online context. This work shows how complex can be the dialogue between the structure of a system and its emergent cooperative behavior but also offers some interesting insights especially about the use of PED in an online high-order interactive environment, such as r\Place.
Comprendere il comportamento cooperativo nei giochi online: un approccio di entropia parziale
COTA, ANDREA PIO MARIA
2022/2023
Abstract
In social phenomena the basic constituents are humans, i.e., complex individuals who interact and cooperate with a limited number of peers compared to the total number of people in the system, with the aim of achieving a common goal. Understanding and modeling collaborative phenomena is important to explain and predict what really happens in the society we live in, in terms of culture, laws, social contexts and much more. More recently, we have also begun to interact with others in an online setting, for example we can share our personal updates on social networking sites, engage with and comment on blog entries, but also participate in virtual communities, such as the case of "Reddit Place", a social experiment created by one of the most famous social networks, Reddit, in 2017. Participants (users) had access to a drawing canvas where they could change the color of one pixel at every fixed time interval. Users were not grouped in teams nor were given any specific goals, yet they organized themselves into a cohesive social structure and collaborated to the creation of a multitude of artworks. The advantage of online social interactions is that they can be easily recorded in the form of dataset and then studied for scientific purposes benefiting research, in fact, in this work we performed a study on a dataset comprising more than 16M user actions, recorded on the online social experiment, with the goal of discovering how people collaborate and organize on online social games. Since collaboration patterns are difficult to capture when the relationships between actors are not directly observable, in this thesis we are going to use information theory tools in order to identify specific patterns of collaboration; one of these is Partial Entropy Decomposition (PED) which allows us to distinguish the behavior of users who cooperate from those who do not. To understand the community dynamics, we, made several definitions of interac- tion between users, some based only on the placement of their pixels others that also included their color. Moreover, thanks to the timelapse of the experiment, we noticed how different artworks had different behaviours during the whole time: some of these were disappearing, while others remained in the same positions, probably indicating the presence of a conflict, in the former case, and peaceful relations, in the latter. Finally, we attempted to replicate the dynamics of the canvas on which users inter- act by generating synthetic timeseries using Gillespie’s algorithm and our definitions of interactions, comparing the various results with the entropy measures mentioned earlier showing how PED is an extremely useful tool to identify collaboration patterns also in an online context. This work shows how complex can be the dialogue between the structure of a system and its emergent cooperative behavior but also offers some interesting insights especially about the use of PED in an online high-order interactive environment, such as r\Place.File | Dimensione | Formato | |
---|---|---|---|
866282_tesifinal.pdf
non disponibili
Tipologia:
Altro materiale allegato
Dimensione
5.65 MB
Formato
Adobe PDF
|
5.65 MB | Adobe PDF |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.14240/146323