3 Things We Can Learn from Sociolinguistics
As qualitative and mixed-methods researchers, we all know that words hold powerful insights. But do we understand just how powerful speech can be? This week we dive into that question with Dr. Nicole Holliday, Assistant Professor of Linguistics at Pomona College.
Dr. Holliday is a sociolinguist. Sociolinguistics is the systematic study of social variation in language—how accent, dialect, word choice, intonation, and other language choices are used by people as tools to navigate the social world. Dr. Holliday is interested in how people use linguistic variation to construct their social identities and understand the identities of others.
In this episode, we dive deep into the study of language and its impacts on the social world. We touch on social judgements and language bias, both in person and in the technological space. We also talk all about language bias in the tech and research world, and how we can work to avoid (or at least be aware of) biases in our own research.
Here are some of the things that I learned with Dr. Holliday:
1. We all make judgments based on language
This probably isn’t news to most of us. But one space we might not think about implicit bias is in our perceptions of other peoples’ speech.
When we hear someone talk—even when we can’t see their face—we pick up on speech patterns that give us a social impression of the speaker. That includes their age, their ethnicity, their gender, and many other social variables.
These judgments can be made based on very small, almost imperceptible differences in speech. The sound of a vowel, or a slight pitch change in voice, can be enough to trigger judgment.
Dr. Holliday has seen through her research (and the research of many other linguists) that we are making speech judgements even when we think we aren’t. As qualitative researchers who deal primarily with spoken and written language as data, it’s important to keep in mind that we are likely making judgment calls about our participants, whether we’re fully cognizant of it or not.
2. Be honest about your goals and expectations
Dealing with research bias (linguistic or otherwise) can be extremely difficult. It can be all the more difficult to eliminate biases in research when you are passionate about what you study, or share a social relationship with those you are learning about.
Dr. Holliday suggests that one reason for this is that we all have ideas coming into research about what we expect to see. Even in generative research, many of us have implicit “hypotheses” about what we might find. In industries, our stakeholders may even have explicit hopes or expectations for a particular study.
In these cases, Dr. Holliday says that getting rid of that expectation might be unrealistic. But one thing we can do is at least be transparent about the expectations we have. Acknowledge at the outset of a report who you are, what stake you have in the project, and what expectations you and your stakeholders had going in. This is a good way to show your cards and let people take those pieces of information into consideration when reading your results.
"We always do a project expecting that we're going to find something one way or another, or even if we're not explicitly doing hypothesis testing, we have a feeling. You should be honest about that with yourself and with your collaborators, maybe even in your write-up: we expected to find this, we actually found this other thing."
Dr. Nicole Holliday
Assistant Professor of Linguistics, Pomona College
3. Design—and research—with variation in mind
In this episode, Dr. Holliday tells us all about her recent work with Automatic Speech Recognition (ASR). She, along with other language scholars, have uncovered multiple biases inherent in ASR technology.
In her work, she has found that most ASR programs are much worse at transcribing speakers who speak English as a second language than they are at transcribing native speakers. This is a huge issue, since the vast majority of people who speak English worldwide are nonnative speakers.
Dr. Holliday’s team was able to dig deeper into this bias and find that even within the ESL group, there was variation in who was able to be transcribed, depending on the first language they spoke. She argues that there are alternate models that might be able to adapt to these speakers better, if there was a way to discern and address this variation within the product experience.
Her story underscores the importance of building products that expect and account for the wide variety of users out there, linguistic or otherwise.
Interested in checking out other People Nerds podcast episodes? Read more of our breakdowns here.
Karen is a researcher at dscout. She has a master’s degree in linguistics and loves learning about how people communicate with each other. Her specialty is in gender representation in children’s media, and she’ll talk your ear off about Disney Princesses if given half the chance.