Eye on AI - July 12th, 2019
Welcome to Aigora's "Eye on AI" series, where we round up exciting news at the intersection of consumer science and artificial intelligence!
This week we cover three main topics: the quantification of experience, how text analytics are producing new insights, and the interaction between technological advances and retail. Enjoy!
Quantification of Experience
We begin with week with exciting news out of The Sociable, given the long-standing interest within sensory in predicting sensory profiles based on instrumental measurements. According to the recently published article, “Scent-ient machines are making sense of smell and taste,” companies like Silicon Valley startup Aromyx have begun digitizing human and animal sensory reactions – smell, taste, etc. – to predict human and animal reactions to foods, aromas, chemicals and more.
“By using the human receptors themselves,” Chris Hanson, Aromyx CEO, writes, “Aromyx can achieve objective measurements of taste and smell which should give a much more accurate map of human perception.”
Application is practically universal, with particular interest from the food and beverage, consumer packaged goods, chemicals and agriculture, automotive, and defense and aerospace industries.
To better understand how Aromyx digitizes smell and taste on their “EssenceChip” system, please watch this short video:
Next, a new study by Dartmouth College, published in the Proceedings of the Royal Society B, found that sounds, shapes, speech and body movements convey the same type of emotional arousal across the senses, helping to explain why nearly anything can have emotional tone and how the same emotional language can be used to describe seemingly unrelated experiences.
“... people automatically perceive the frequency spectrum of whatever is coming into their ears and eyes and compute the average," explains senior author Thalia Wheatley, principal investigator of the Dartmouth Social Systems Laboratory. "This is how people quickly identify the amount of emotional arousal in a person's voices and movements but also in abstract shapes and sounds, such as why spiky shapes seem to convey higher arousal than rounded shapes."
This research could have a downstream impact on the understanding of emotion within consumer communications, so it's worth keeping an eye on progress in this area.
Text Analytics Producing New Insights
Text data are one of the sources of data that are widely available in large quantities to consumer scientists, so we now turn to the topic of how machine learning algorithms are producing new insights when applied to large-scale text corpora.
First, using a machine learning algorithm called Word2Vec, researchers from Lawrence Berkeley National Laboratory were able to sift through millions of abstracts from materials science to make new discoveries that were previously missed by humans.
“It can read any paper on material science, so can make connections that no scientists could,” says researcher Anubhav Jain. “Sometimes it does what a researcher would do; other times it makes these cross-discipline associations.”
The algorithm teaches itself to understand complex concepts associated with the targeted subject, then links words associated with those concepts to papers previously used for research not linked to the targeted topic. This allows researchers to uncover new information on targeted subject matter from materials that had previously been collecting dust in the archives.
“You could use this for things like medical research or drug discovery,” says Vahe Tshitoyan, the lead author on the study. “The information is out there. We just haven’t made these connections yet because you can’t read every article.”
In one experiment, researchers used digitized papers published before 2009 to predict one of the best modern-day thermoelectric materials four years before it was uncovered in 2012. While there is not the same volume of abstracts available to sensory and consumer scientists as there is to material scientists, I would be very interested to see these techniques applied to the abstracts of Food Quality and Preference and the Journal of Sensory Studies. If you are likewise interested in such an application, contact me and let's talk!
Moving on to another excellent application of text analytics, Monell Chemical Senses Center recently used machine-learning to sort through thousands of online food reviews, posted on over sixty thousand products, to try and identify people’s chief taste complaint. What they discovered is that people’s primary taste complaint was too much sweet. Words associated with sweetness were routinely identified in negative reviews.
Due to the over-sweetened market, might ‘sweet’ soon be a taste of the past?
Technology and Retail
We wrap up with a look at the interaction between the technology and retail sectors. While it’s no secret that technology has played a large role in brick and mortar retail stores being on the ‘endangered’ list, one way to revive them could be through AI and other adjacent technologies, as reviewed by Frank Cittadino in Forbes.
And finally, as reported in Venturebeat, chatbots continue to make inroads in retail, this time in helping customers to place pizza orders quickly. Inference Solutions, a San Francisco-based company that develops virtual agents for sales and services, announced this week that it has partnered with Pizza Hut Australia to implement a chatbot-based service that will route calls to representatives, nearby Pizza Hut locations, or take orders, learning all the while to make more educated decisions.
“The ability to easily add new areas of automation will enable Pizza Hut Australia to continue to improve customer experience while reducing their cost-to-service,” said Inference Solutions CEO Callan Schebella.”
Similar chatbot applications are beginning to show an uptick worldwide, with research groups predicting that conversational assistants will contribute eight billion dollars annually in cost savings by 2022.
Other News
A reminder that videos of consumer interviews are themselves data
People can now donate phone compute-time while they sleep to help identify anti-cancer compounds
Acknowledgement
We thank Langdon Moss for his contributions to this article!
That's it for now. If you'd like to receive email updates from Aigora, including weekly video recaps of our blog activity, click on the button below to join our email list. Thanks for stopping by!
