I conduct research on natural language meaning: currently, I focus on ad hoc implicatures, perspective-taking and informativeness. Specifically, I investigate the way these processes intersect during language communication using simple and fun referential games. To answer my research questions, I make use of methods and models from theoretical linguistics, computational psycholinguistics and neuroscience.

Methods

I collect data through a variety of methods:

• production and comprehension experiments, in the laboratory or via the internet
• psychological techniques (eye-tracking)
• neuroscientific techniques (EEG)
• computational cognitive modelling

Languages

To understand cross-linguistic variation and how different processes unfold in language communication, it is vital to study languages beyond English. For this reason, my research also compares how pragmatic reasoning, perspective-taking and informativeness interact during language communication in English, Greek, and Chinese, focusing on both neurotypical and neurodiverse (primarily autistic) populations.

Current Research Areas

Ad Hoc Implicatures

When a speaker says “Give me the bag with apples”, and there are two bags—one with only apples and another with apples and oranges—the listener understands that the speaker wants the bag containing only apples. The listener reaches this interpretation by deriving an ad hoc implicature using the visual context. However, if the same expression is used when there is only a bag with apples and oranges, the listener needs to understand that the speaker is referring to this bag. In both instances, arriving at the correct interpretation requires pragmatic reasoning. My research explores the cognitive processes behind this type of pragmatic reasoning cross-linguistically in both neurotypical and autistic speakers and listeners.

Visual Perspective Taking

In a classroom, a teacher stands in front of a map with a mountain on the left and a river on the right from their perspective. The students, facing the teacher, see the mountain on their right. When the teacher asks, “Can you point to the mountain?” a student who understands the teacher’s perspective will point to the left side of the map, even though it appears on their right. This process is called Visual Perspective Taking. My research investigates how Visual Perspective Taking influences speakers’ expressions in tasks requiring pragmatic processing and the corresponding effects on listeners’ comprehension. We create novel tasks, building on previous research, to explore the effects of visual asymmetries between neurotypical and autistic speakers and listeners on pragmatic reasoning.

Informativeness

Many studies on referential communication have shown that speakers tend to be over-specific in their instructions; that is, they provide more information than necessary for the listener to identify the referential object. Here, we capitalise on this finding to investigate the effects of over-specificity on pragmatic reasoning.

Autism

Enhanced Rationality has recently been proposed as a new framework for research on autism. Traditionally, most studies on autism have focused on identifying difficulties, rather than on exploring the skills in which autistic individuals might excel compared to neurotypical individuals. In contrast to this focus on difficulties, my research investigates whether there are differences between autistic and neurotypical speakers/listeners in language communication tasks, in line with the concept of enhanced rationality. The goal is to identify the strategies that different neurotypes may use, rather than to pinpoint areas of difficulty. Understanding these strategies could have significant clinical implications, potentially informing tailored communication approaches and interventions that leverage the strengths of autistic individuals.

Large Language Models (LLMs) and Social Reasoning

Open-source LLMs can help us understand social reasoning and language communication across different neurotypes. These models provide access to their training data, architecture, and procedures, allowing for modifications that facilitate the investigation of cognitive processes underlying social reasoning. This approach could also shed light on the difficulties autistic and other neurodiverse individuals face with pragmatic reasoning, leading to targeted interventions to support them. To this end, I am testing state-of-the-art LLMs capable of integrating visual and/or linguistic information to interpret indirect expressions in the same communication games I run with human participants. The goal is to better understand how these processes unfold in human cognition, when they go awry, and the steps required to address these difficulties in LLMs, with the ultimate aim of developing practices that could also benefit human individuals.