this essay uses conversation-design as a crutch to talk about design’s identity in a world defined by engineering’s obsession with ‘low-hanging fruits’ and human-like interfaces.
when google publicised duplex in 2018, most people were terribly impressed by how natural the computer-generated voice sounded. there has been some controversy1, of-course, but it is no match for the sheer power with which google (and big-tech) continue to massage such technologies into our lives.
while the work done by big-tech in replicating human-ness is impressive, impressive does not always mean appropriate or timely or socio-culturally useful. (i can be overly-dramatic and allude to the sheer impressiveness of the nuclear bomb ; let me not digress, however.)
design: growing popular yet marginalised
conversational technologies are still in their infancy: engines still take several seconds to understand words and speak back to us, agents struggle with context-switches and cultural references, and software are still quite ignorant of situated real-world connections. the development of these technologies is engineer-led ; for all the hype surrounding the rise of conversation design, designers (and their ilk) remain fringe-players in the development process — working mostly to make interfaces appear more natural than they actually are.
things continue to evolve, though: we are gradually beginning to engineer (and design) experiences that are less discrete than “set an alarm for 8 am tomorrow” or “will it rain today?” ; conversations with machines can now last several turns (or, if you’re interacting with woebot2, several days), thread together several touch-points, and mediate complex experiences3 ; and conversational technologies grow a lot more sophisticated, immersive and multi-modal with every passing month.
conversational technologies that power duplex are here to stay — and they should! — they’re immensely powerful, and promise us gifts of mind-numbing convenience. engineers and algorithms — who’re are put to work on google duplex, samsung neon4, hanson’s sophia, amazon alexa, apple siri, magic-leap’s mica, etc — have grown to master the art of ‘mechanising high-fidelity human-likeness’ and, with minimal design-input, are able to build conversational interfaces that appear incredibly natural. that is what makes me worry: what is design left with?
engineered technologies and designed interfaces
i think it helps if we look at the previous paragraph and recognise that a technology is not the same as an interface. this is how i see it: google’s ranking-and-sorting algorithms are a technology ; its website and voice-assistant are two interfaces (that allow me to interact with google’s technologies) ; and google-search itself serves as a medium.
a single technology may manifest through any number of interfaces: to submit a search-query to google, i may use either a browser or my voice. on the other hand, a single interface may also allow any number of different technologies to flow through it: a voice interface lets me interact with maps, set reminders, control music, dim lights, make phone calls, and so much more.
technologies and interfaces are tightly interwoven ; and neither should be allowed to dominate the other. interfaces must not be considered secondary to the technologies they help manifest ; and a designer should be allowed more control over how human-like, if at all, an interface is.
low-hanging fruits: a double-edged sword
people tend to be suspicious of unknown things, and may reflexively flinch away from outstanding products without trying them ; that is why engineers like to clothe their technologies in human-like interfaces ; cloaking newness under a garb of familiarity is an obvious way to overcome this problem, and is usually quite easy to do. (since easy is often quick, this habit seduces investors too.)
the tech-world’s ‘easy’ approaches to socially-dispersed products, however, are often morally questionable: after-all, we live in an age where “social” technologies have promoted loneliness, depression, misinformation, ignorance and polarisation on an impressively global scale.
the trouble with human-like interfaces
humans don’t just interact : we relate. a human’s appearance and behaviour encodes her situatedness, her empathy, her understanding of complex existential notions, her ability to respond humanely to non-historical patterns, and her unquestionable potential to form relationships. a machine lacks this ability ; so why must a human–machine interaction ape a human–human relationship?
first-impressions make up a very small fraction of our total experience with a technology, but disproportionately
influence taint its future development. human-likeness sets inaccurate expectations in the mind of the audience ; it can deflect or amplify a technology’s psychological, socio-cultural and ecological impact ; and the messaging can also bite its own tail by subverting the medium and stunting its growth.
this is where i think design should play a role: in demonstrating the pitfalls of this approach, and crafting ways to make technology seem less daunting without mindlessly infusing it with human-likeness. so, instead of submissively gilding engineered products for sale, designers should proactively challenge5 the unsolicited development of features like sarcastic dialogue, feminine/submissive persona, filled-pauses and breathing sounds in conversational agents.
for our good and their own, today’s machines must not be allowed to set false expectations: new mediums invite new behaviours, and humans — who’re better at adapting to changing landscapes than engineers give them credit for — can be coaxed to embrace new technologies through appropriately designed interfaces.
or else: we’re only being dogmatic.6
human-like and inhumane
humane is a buzz-word in the industry today ; i’m all for it, as long as it isn’t conflated with human-like. humane is necessary ; mindlessly human-like is problematic: inconsiderate, dishonest, immoral and ultimately inhumane.
the criticism sounded something like this: “if you know that it’s a machine you’re on phone with, you may get creeped-out by backchannels like ‘uh-huh’ or artificial breathing sounds ; if you don’t know, however, and find out after a call (that you’d just spoken with a machine), you may feel cheated.” — i wonder, though: if duplex helps automated-spam-calls become an unmanageably large nuisance, will people grow less addicted to their “always connected” lifestyles and look up at the sky more often? i think that would be nice. ↩
we live in an age where machines mediate human relationships: these relationships can be human–human, human–environmental, cultural–human, or some other combination. when seen this way, industrialised notions of human–machine interaction design do feel a bit narrow. ↩
earlier in 2020, samsung boasted that neon’s core technology can “autonomously create new expressions” and “new movements” that are “completely different from original captured data”. i only ask: in an age where technologies are failing to mediate the rich repertiore of emotions we already have, why do we fetishise an algorithm’s ability to invent new ones? ↩
this ends-up being less about design, and more about corporate power-play. ↩
i’m not a fan of the church of big-tech. it stinks. ↩