Are Voice-Activated Chatbots Really the Future of A2J Tech?

Subtitle: 🤷🏽‍

There have been a couple of articles out this week about the future of voice-activated chatbots used in the legal aid context. While I think the concept is really interesting, I’m feeling skeptical about the breadth of its application.

Inclusive Design

A good reason for looking at voice-activated bots is accessibility. So to start, a wee bit of background for the uninitiated. Inclusive design (also sometimes conflated with universal design and accessible design) is the design of digital products (or anything, really) where the needs of users with disabilities are specifically considered. This can mean ensuring the product is readable by screen readers, making sure the user can navigate using a keyboard, providing sufficient color contrast, not using color as the only way to tell the user something, etc. These design elements are critical for the users that need them, and do nothing to impede the experience of users who don’t need those accessibility features. Indeed, good universal design should make for a better experience even for users who do not have disabilities.

An oft-cited example from the physical world are curb ramps. Curb ramps became a requirement under the Americans with Disabilities Act specifically to allow people with disabilities to be able to have universal access to sidewalks, but this feature has also had the consequence of helping parents with children in strollers, the elderly, or people on Segways. (Just kidding, Segways are the worst and they should stay off sidewalks. And really everywhere else).

The use of voice-recognition applications have been experimented with for a while to provide access to information for users with disabilities. My favorite is the super creepy Nadia project from Australia (who uses the voice of Kate Blanchett — seriously, she apparently has a universally soothing voice). Nadia is next-level stuff. She can understand user emotions by reading facial expressions via camera. Notably, though, Nadia was designed specifically for use by users with disabilities, which should be distinguished from products that are designed for universal use that includes users with disabilities. It will be interesting to see how this type of product evolves and scales for all users and what use cases will be successful.

Where does Voice Recognition Fit In?

Like anything else in product development, selecting a voice-activated chatbot interface is a design decision that needs to be made in the context of the application’s goals and who its users are. And like all interface choices, there are tradeoffs that need to be considered.

A voice-activated chatbot can be extremely powerful where a user needs quick and brief information, especially where the user either has physical or attentional limitations (either due to a disability or because they are typically carrying something or driving when the need for the bot arises). Also, a voice-activated bot makes sense where the input the user needs to provide to the bot for the bot to respond with useful information is also brief and simple. If these boxes aren’t checked for a particular use case, I’m not sure I see the rationale behind using voice recognition bots.

Take these two scenarios in the legal aid context. Imagine that a court has a voice recognition bot for visitors of the court to use. The user can ask basic questions about the courthouse, like “where can I park at the court?” or “how early should I show up prior to my hearing.” This seems like a great use case for such an interface — brief, easy exchanges that get the user information fast.

In a second scenario, imagine a bot that addresses a substantive area of law, like employment. Sure, there are some things that can be easily shared back and forth in such an interface — “can an employer discriminate against me because of my race” — but once we get any deeper, such as “help me file a complaint with the EEOC” or “what are my rights at work regarding overtime pay?”, we begin to run up against the limitations of this interface.

The law is complex. I know that we’re trying to make it less complex using technology, which is a fine idea. But often times a major part of that job is to make sure that the law is digestible and understandable. I shudder at the idea of a robot speaking for more than a sentence at a time about legal rights and obligations. That seems like the wrong interface for delivering anything more than the absolute basic legal concepts to a consumer.

I don’t think the accessibility angle tips the scale here either. A well-designed visual interface (as opposed to voice interface) will allow for people with disabilities to be able to digest information in whatever way they need, while allowing for substantially greater design flexibility. Also, I giggle a little when I hear that these interfaces are good for the elderly. My dad’s texts via Siri are just absurd…

I, admittedly, have never tested a voice-activated chatbot with a user in the legal context (although I have done user testing sessions with low-income individuals in need of legal services with more traditional chatbot interfaces). I’m all for getting some prototypes out there and am curious to see the results. I just worry any time I see a lot of excitement about a technology in legal tech (blockchain! AI!) that people are thinking of technology over the real need for the product or the optimal interface for the user.

So, I’m curious to hear from anyone who has tested these in the legal context — what have you seen? What are the use cases that you’ve found to be successful? Am I lacking in imagination about the possibilities here? Would love to hear any thoughts in the comments below.