Research Spotlight
This webpage is dedicated to comments and brief explanations related to the research collaborations and initiatives promoted by Complex Science Consulting.
lexic
Multiplex lexical networks combine semantic and phonological similarities of words.
lexic2
Healthy people can name words of high/low closeness with the same probability. People with aphasia can name words of high closeness considerably better than words with low closeness.
Understanding aphasia with multiplex lexical networks
Castro N. and Stella M., Journal of Complex Networks (2019) https://doi.org/10.1093/comnet/cnz012.
Aphasia is a cognitive pathology progressively disrupting speech, writing and language understanding.
Picture naming involves language understanding and production. It is an experiment where an individual sees a specific picture and should say the word indicated by the pic. People with aphasia have issues in recalling words and using language correctly, so that to them picture naming might go wrong quite frequently.
Our main achievement in (Castro and Stella 2019) was identifying a “production gap” in people with aphasia, who find words more or less difficult according not only to standard psycholinguistic features of words (e.g. length, frequency) but also and especially according to the location of these words in the mapping of semantic and phonological word similarities. Imagine a network of roads in a city, some locations might be more difficult to reach by car than others.
We found that the same applies to words in picture naming experiments: people with different forms of aphasia have a considerably higher chance of producing mistakes in naming words that are “difficult to reach”.
Important note: the difficulty arises only when phonology and semantics are combined together in a multiplex lexical network (see Stella et al. 2018). Individually, semantic layers and phonological layers do not highlight this production gap.
This preliminary study opens important new challenges for using multiplex lexical networks in the investigation of clinical phenomena related to cognition.
games2
games-10-00016-g001
Modelling polarization with multiplex lexical networks
Lipari F., M. Stella and Antonioni A., Games (2019) https://www.mdpi.com/2073-4336/10/2/16/htm.
The adoption of an action can represent the diffusion of a wide variety of things, such as trends, fashion, ideas, political beliefs among many others. In our model (cfr. Lipari et al. 2019), the adoption of an action happens on a social network where individuals are connected together but only at a local scale. Social links can be severed and rewired by individuals in order for them to connect with others having the same adopted action. Actions can be imitated by individuals looking what most of their linked neighbours/friends do. These actions diffuse over a layer of interactions we call Evident Layer.
The strategy of cutting or preserving links can be learned from other individuals according to a separate layer of interaction we call Hidden Layer. This layer might represent social contexts different from the friendship interactions, e.g. social ties from an office environment or religious interactions and so on.
Our model indicates that when coupled together, the dynamics of action adoption of the evident layer crucially depends on the arrangement of social ties in the hidden layer, where strategy learning happens.
This theoretical finding is important because in some instances the specific arrangement of the hiddel layer can lead to strong polarisations in the evident layer.
Think about the polarisation of voting during massive elections. Is this polarisation due just to the imitation of actions during voting or might there be hidden layers of additional interactions from different social/cognitive contexts that actually drive voters apart?
Further research towards this direction is required, as indicated by this study.