Towards Ecologically Valid Research on Language User Interfaces

By Harm de Vries, Dzmitry Bahdanau, Christopher D. Manning. 2020

61 citations recorded by the source corpus.

Language User Interfaces (LUIs) could improve human-machine interaction for a wide variety of tasks, such as playing music, getting insights from databases, or instructing domestic robots. In contrast to traditional hand-crafted approaches, recent work attempts to build LUIs in a data-driven way using modern deep learning methods. To satisfy the data needs of such learning algorithms, researchers have constructed benchmarks that emphasize the quantity of collected data at the cost of its naturalness and relevance to real-world LUI use cases. As a consequence, research findings on such benchmarks might not be relevant for developing practical LUIs. The goal of this paper is to bootstrap the discussion around this issue, which we refer to as the benchmarks' low ecological validity. To this end, we describe what we deem an ideal methodology for machine learning research on LUIs and categorize five common ways in which recent benchmarks deviate from it. We give concrete examples of the five kinds of deviations and their consequences. Lastly, we offer a number of recommendations as to how to increase the ecological validity of machine learning research on LUIs.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Natural Language Processing, Human-Computer Interaction

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Natural Language Processing

In Human-Computer Interaction

Citations in the corpus

UnifiedSKG: Unifying and Multi-Tasking Structured Knowledge Grounding with Text-to-Text Language Models (2022, background): “Furthermore, training and evaluation data should reﬂect the intents and linguistic phenomena in the real world (de Vries et al., 2020), suggesting more realistic tasks to be added into U NIFIED SKG.”
Dynabench: Rethinking Benchmarking in NLP (2021, background): “While crowdsourcing has been a boon for large-scale NLP dataset creation (Snow et al., 2008; Munro et al., 2010), we ultimately want NLP systems to handle “natural” data (Kwiatkowski et al., 2019) and be “ecologically valid” (de Vries et al., 2020).”
Deploying Lifelong Open-Domain Dialogue Learning (2020, background): “Secondly, as crowdworkers are motivated by pay, not by interest in the actual tasks themselves, the data distribution may not match the desired one (de Vries et al., 2020).”
Formalizing Trust in Artificial Intelligence: Prerequisites, Causes and Goals of Human Trust in AI (2020, background): “ome specific way, for which gold labels are available. Test sets are generally seen as imperfect approximators of post-deployment data, as their distribution is not representative of true unseen data [12] and thus cannot imply overall generalization. However, this perspective only takes into account”
DynaSent: A Dynamic Benchmark for Sentiment Analysis (2020, methodology): “First, by harvesting naturally occurring examples for Round 1, we tap into a wider population than we can via crowdsourcing, and we bring in sentences that were created for naturalistic reasons (de Vries et al., 2020), rather than the more artiﬁcial goals present during crowdsourcing.”
Model Alignment as Prospect Theoretic Optimization (2024, background): “Ecologically valid evaluation (De Vries et al., 2020), where the aligned models are deployed in real-world settings, are also needed to judge the merits of different HALOs.”
Conversational Assistants and Gender Stereotypes: Public Perceptions and Desiderata for Voice Personas (2020, methodology): “Methodology: We explore an alternative methodology to the usual way we get ‘users’ involved in NLP research, i.e. crowd-sourcing or scraping online data (de Vries et al., 2020) borrowed from HCI.”
Evaluation Gaps in Machine Learning Practice (2022, background): “On the other hand, application-centric evaluations are concerned with how the model will operate within an ecosystem consisting of both human agents and technical components (Figure 1), sometimes described as the “ecological validity” [46].”
Can Small and Synthetic Benchmarks Drive Modeling Innovation? A Retrospective Study of Question Answering Modeling Approaches (2021, background): “Datasets that more closely resemble natural language are superior to those that to do not (Manning, 2006; Kwiatkowski et al., 2019; de Vries et al., 2020).”
Dialogue in the Wild: Learning from a Deployed Role-Playing Game with Humans and Bots (2021, background): “Secondly, as crowdworkers are motivated by pay, not by interest in the tasks themselves, the data distribution may not match the desired one (de Vries et al., 2020).”