Social Norms Guide Reference Resolution

Abstract

Humans use natural language, vision, and context to resolve referents in their environment. While some situated reference resolution is trivial, ambiguous cases arise when the language is underspecified or there are multiple candidate referents. This study investigates how pragmatic modulators external to the linguistic content are critical for the correct interpretation of referents in these scenarios. In particular, we demonstrate in a human subjects experiment how the social norms applicable in the given context influence the interpretation of referring expressions. Additionally, we highlight how current coreference tools in natural language processing fail to handle these ambiguous cases. We also briefly discuss the implications of this work for assistive robots which will routinely need to resolve referents in their environment.

Norm Referenced Test Sebagai Metode Penilaian Berdasarkan Kelompok Kelas

“…Pengenalan Norm Referenced Test atau dalam bahasa lokal disebut dengan Pendekatan acuan normatif atau penilaian acuan normative adalah kerangka kerja baru yang memetakan perbedaan individu terhadap model referensi, memungkinkan inferensi statistik pada tingkat individu dengan memetakan persentil variasi di seluruh populasi (Rutherford et al, 2022), Pendekatan ini sangat berharga di bidang seperti psikiatri komputasional dan ilmu saraf klinis, menawarkan solusi untuk beralih dari perbandingan tradisional antara pasien dan kontrol sehat menuju pemahaman yang lebih mendalam tentang heterogenitas biologis dalam gangguan mental (Abrams & Scheutz, 2022). Pemodelan normatif menyediakan protokol standar untuk menganalisis data, membimbing pengguna melalui pemilihan input, pilihan pemodelan, dan analisis lanjutan seperti stratifikasi individu berisiko tinggi dan pemodelan prediktif perilaku (Kanjee, 2010).…”

Culturally Aware Natural Language Inference

“…Our work is also related to reasoning about social norms. Social norm reasoning tasks involve social judgement (Forbes et al, 2020;Shen et al, 2022;Fung et al, 2022;, intents and consequences (Emelin et al, 2021;Rashkin et al, 2018), and reference resolution (Abrams and Scheutz, 2022). A recent line of work has focused on the situational and defeasible nature of social norm reasoning (Rudinger et al, 2020;Pyatkin et al, 2023).…”

EtiCor: Corpus for Analyzing LLMs for Etiquettes

“…The authors create a new dataset, Commonsense Norm Bank, that incorporates existing datasets such as Social Chemistry (Forbes et al, 2020), Scruples (Lourie et al, 2020), Ethics (Hendrycks et al, 2020), Commonsense Morality (Frederick, 2009), Moral Stories (Emelin et al, 2021), and Social Bias Inference Corpus (Sap et al, 2020). Delphi uses a pre-trained language model Abrams and Scheutz (2022) examines how pragmatic cues (in the form of social norms) affect the resolution of mentions in the task of conference resolution. In the study, different daily life situations are textually described, and the machine response is compared to the human response, giving an idea about how much social context is utilized by the model.…”