2040’s Ideas and Innovations Newsletter, Issue 65: Behavioral and Inference Data, A 360 Perspective
What You Say Is Very Different From What You Do and How You Act
Issue 65: July 21, 2022
What we say or how we respond publicly is often very different than what we may feel, believe, like or dislike privately. Let’s consider a hypothetical conversation with a friend. She is proudly wearing a new dress that she just purchased. She asks your opinion as she seeks confirmation if the dress makes her look good. She surely thinks it does. You of course don’t want to hurt her feelings. You express that you really like what she is wearing, even though you think the pattern and color look more like a tablecloth or set of curtains. And this may be even though the style is totally on trend; you just can’t get beyond your first reaction.
Here’s a work-related example. In discussion with your work team, you are bored with the items being reviewed, and your mind begins to drift off. You then hear your name and call yourself back to attention. Your teammates are asking if you agree with the approach they want to take to solve the problem. You quickly nod your head in agreement, despite not really understanding what the team wants to do. Even though you don’t know what has been discussed, the team members depart the meeting believing there is consensus across the team, including you.
Here’s a third example. You receive a survey request that is offering the opportunity to win a gift certificate if you participate. You always feel lucky and usually respond to offers to potentially get something free for a few minutes of your time. You answer the questions presented in the survey even though you don’t care about or relate to the subject matter.
What these situations have in common is that in each instance you responded with information or an opinion that was what the recipient wanted or prompted you to express. And in each instance, what you shared was counter to your own thoughts and beliefs.
Unintended Consequences
The truth is that our responses can lead to future actions or feelings by others that run counter to who we really are, want or like. And there can be unexpected reciprocity. Your friend may give you a gift similar in pattern and/or color to her dress. Your team members put you in charge of the project based on your quick agreement to their plans. And the organization that sent you the survey is now drowning you with emails about content and products related to your responses.
Human nature is often predictable. We don’t like to hurt people’s feelings, don’t like to disagree, and aren’t comfortable being confrontational. We also feel compelled (often indebted if we are getting something in return) to give others what they want, make people feel good, and help them feel confident about their choices.
Depending on our motivation, the value of our direct responses can be very different from who we are, how we feel, what we like, and what we really need. In the context of business, this can become a nightmare for anyone collecting data and creating a personalized customer profile.
What we do combined with what personal information we share can be easily leveraged to assume who we are, what we like, how we feel, what we believe … and so much more. As individuals go site to site, app to app and digital experience to the next, they often don’t understand that they are generating data. Plus, they don’t recognize where that data is stored and how it may be used by others to solicit them now or in the future. In a data-driven world, we can become an imposter or reveal our most intimate personal thoughts and beliefs by the actions we take, consciously or subconsciously.
Data Breadcrumbs
Pre-digital, it was hard to collect the bits and bytes of information necessary to infer who we are, what we may need, be interested in, and believe in. Very little evidence existed in one place that would provide an accurate picture. Consider the primitive methods of broadcast advertising, direct mail campaigns, and cold-calling salespeople using a paper phone book going from number to number. What little data that was available was leveraged generally through spray-and-pray techniques or as shots in the dark. Needless to say, it was an unproductive and inefficient exercise.
Today in our increasingly immersive digital world, our actions over time are easily amassed and analyzed to reveal ourselves and predict what we might read, buy, interact with, or do next. The digital trail of behavioral and other action/information-based data can be used to infer. The concept of inference and inference data is data that in part and sum are brought together, merged with other data, analyzed, and used to determine or predict action. Inference models using data can build a hyperconnected personal profile, or identify who we are (name, location, SSN, family members, etc.) and then leverage that information for financial gain, whether directly or by selling it to a third party.
Nowhere to Hide
Data are used to create inferences across our digital actions and include everything from the content we consume, products we research and buy, the information we share, and the GPS location of our internet connections and mobile devices, to the music, podcasts, shows and videos we listen to or watch. And on and on. Our spending patterns (store preferences, reward programs, credit card usage, health care and medication activities as well as the data that we are compelled to share with the government can be leveraged to connect to other data that eventually erases our anonymity.
So, in today’s data-driven economy, data used by organizations to create inferences, via machine learning, algorithms, or even manually promise to produce liquid gold. But buyer beware: The human programmers creating machines and algorithms invariably instill their own biases and faults, which will result in faulty outputs.
Data Goldmines
Most of us are oblivious as to how recommendations magically appear on our shopping platforms, why certain coupons are delivered by mail, why advertisements for products we are considering follow us across the sites we visit, or why emails are sent with content we are interested in. We appreciate, for the most part, what we perceive as someone or something looking out for us with the benefits of convenience, time-savings, or an opportunity to fill a few moments of boredom.
There are many upsides, of course, for individuals. Organizations can determine who we are, what we want and how to serve us accordingly. The most obvious examples are Alexa reminders for replenishment programs. Organizations can also sell your data to others who offer complementary products or services that you will most likely interact with. You feel happy because you are experiencing the convenience and the fulfillment of your needs and interests. You may even feel grateful as the advertising, email communications, or texts help you to become more knowledgeable about products and services you didn’t even realize you needed or wanted.
An Organizational Perspective
Data, particularly behavioral and connective data that can be used to create inference, remain king for organizations to achieve growth. Data is critical to any business plan, initiative, or strategy to reach prospective customers, retain current customers, grow market share, and remain competitive. Some organizations have adapted well and have learned to amass data, use it to create inferences, and solidify accurate customer profiles to leverage, curate or even sell to others.
The progressive, truly savvy data-oriented organizations recognize that drawing conclusions based only on what customers say and how they respond can lead to less than stellar results. Rather, inferences can be derived and confirmed based on behaviors which lead to excellent and lucrative results.
We promote the criticality and necessity for organizations to focus on a first-party data strategy, recognizing the concentric circles of connected, informative and related data attributes that cohere as valuable, indicative, and usable. In short, the smart strategy is to collect, identify, value, and measure what matters. The days of leveraging third party cookies, acquiring third-party data and lists, or even commingling data between two organizations is coming to an abrupt end. To be a trusted organization, personal consent is the only recourse to build a data store of customers, acquire new customers and access their data.
Downsides to Individuals
There are downsides for individuals as most of us aren’t directly aware of or involved in how our digital footprints across sites, apps and the like are being manipulated. Let’s use Google as an example. Products like Gmail, Google Search, Google Photos, Google Maps and even Google Drive result in the collection of personal data and archiving it within Google’s ecosystem. When all this personal data is aggregated, Google leverages it in a variety of ways. Google obviously knows who you are, what you search for, the photos you store, and your location. Each bit and byte are recorded and saved. Imagine then, someone interested in knowing more about you doesn’t have to ask you directly. They can just ask Google. As it stands now, Google can share your data with a third party or a government entity, both of which can get access to your Google data.
Inferences based on your data trails can have a downside. Let’s consider a few that have a direct impact on an individual:
Identity InferencesIdentity inferences often focus on determining an individual’s identity from a dataset that appears to lack personally identifiable information. The inferences might link data to a particular name, or they might show how a limited amount of data can be used to re-identify or uniquely identify an individual (Privacy International).
Personal Information InferencesData fed into or used by algorithms, machine learning and AI is leveraged by statistical models to infer all kinds of personal data (name, information provided on forms, address, etc.).
Emotional InferencesAn analysis of the sum of behaviors (likes, loves, post frequency, types of posts, and even word choices made in posts, tweets, or the like) can reveal an individual’s emotional state.
Creepy Data
Let’s look at some study-based examples of how easily what we do exposes who we are. The following selection comes from a compilation Privacy International has available on their website. The choices represent studies that were conducted in 2010-2015 and are still relevant. The intent here is to show you that the collection and use of data to derive inferences has been active for a considerable amount of time.
Are you paying more than others?
“We ran controlled experiments to investigate what features e-commerce personalization algorithms take into account when shaping content. We found cases of sites altering results based on the user’s OS/browser, account on the site, and history of clicked/purchased products. We also observed two travel sites conducting A/B tests that steer users towards more expensive hotel reservations (Hannak, A., et al. (2014).”
Are you revealing more about yourself than you think?
“Showing how user demographic traits such as age and gender, and even political and religious views can be efficiently and accurately inferred based on their search query histories (Bi, B., et al. (2013).”
Are you being classified into a personality type?
“From the analysis, we show that aggregated features obtained from smartphone usage data can be indicators of the Big-Five personality traits [extraversion, agreeableness, conscientiousness, neuroticism, and openness to experience]. Additionally, we develop an automatic method to infer the personality type of a user based on cellphone usage using supervised learning. We show that our method performs significantly above chance and up to 75.9% accuracy (Chittaranjan, G., et al. (2011).”
How are you really feeling?
“In an experiment with people who use Facebook, we test whether emotional contagion occurs outside of in-person interaction among individuals by reducing the amount of emotional content in the News Feed. When positive expressions were reduced, people produced fewer positive posts and more negative posts; when negative expressions were reduced, the opposite pattern occurred. These results indicate that emotions expressed by others on Facebook influence our own emotions, constituting experimental evidence for massive-scale contagion via social networks (Kramer, A.D., et al. (2014)”
Government Action
Governments around the world are proactively recognizing the necessity to enact privacy and data regulations to protect their citizens from organizations that have learned to take advantage of data. GDPR, and now the Digital Services Act and Digital Markets Act in the European Union, seek to protect users from each other and inform users about consent to organizations. Consent is not just for collected or used surface data, but any data generated by a user that can be leveraged to create an inference.
For the public, this majorly manifests in the consent to stop receiving targeted and retargeted ads that leverage data and create inferences for what a user might be interested in or interact with. It also specifically focuses on requiring consent to use all types of data generated or shared by the user to “infer” who the user is and what they want/like to build a profile, which can be leveraged directly or used to amend their profiles held by other organizations, including data broker companies.
All regulation in the EU requires permission first (opt-in), whereas US state-based regulation majorly takes the route of opt-out. Permission is granted until an organization is told to stop. California, Colorado, and Virginia (each with Privacy and Data Regulations coming into force January 2023) address inference data.
VCDPA provides end-users in Virginia an opt-out of the sale of their personal information, and specifically an opt-out of targeted advertisement and data profiling (the collection of personal data and inferences made for the purpose of predicting user behavior). Draft legislation in the US House of Representatives takes a similar approach and defines various types of data that can be used to create inferences. The legislation provides individuals the ability, which organizations must fully and clearly demonstrate, to opt-out of the collection, curation, maintenance, or use of such data. This includes all data that is used in advertising, retargeted advertising, or automated marketing.
Shifting Sands
Individuals can be optimistic that some control will be exerted on what personal data is collected, curated, and used. For organizations which have come to rely on data and the creation of inferences, the sands are shifting. Clarification of what is necessary to gain consent and what types of data that relates to, including deriving inferences, requires some deep thought and consideration of the various state and country-based privacy laws and data regulations being put in place.
Some portion of the user audience will continue to simply agree to having their data collected. Wearing a marketing hat, that is a good thing. However, in the EU, organizations are currently seeing a 50% reduction in data as individuals decline consent. Is the remaining 50% indicative of an entire market of current or prospective customers? That may make inferring from the remaining data more of a guessing game than having confidence in output and outcomes.
References:
Hannak, A., Soeller, G., Lazer, D., Mislove, A. and Wilson, C., 2014, November. Measuring price discrimination and steering on e-commerce websites. In Proceedings of the 2014 conference on internet measurement conference (pp. 305-318). ACM. http://dl.acm.org/citation.cfm?id=2663744
Bi, B., Shokouhi, M., Kosinski, M. and Graepel, T., 2013, May. Inferring the demographics of search users: Social data meets search queries. In Proceedings of the 22nd international conference on World Wide Web (pp. 131-140). ACM. http://dl.acm.org/citation.cfm?id=2488401
Chittaranjan, G., Blom, J. and Gatica-Perez, D., 2011, June. Who’s who with big-five: Analyzing and classifying personality traits with smartphones. In Wearable Computers (ISWC), 2011 15th Annual International Symposium on (pp. 29-36). IEEE. https://infoscience.epfl.ch/record/192371/files/Chittaranjan_ISWC11_2011.pdf
Kramer, A.D., Guillory, J.E. and Hancock, J.T., 2014. Experimental evidence of massive-scale emotional contagion through social networks. Proceedings of the National Academy of Sciences, 111(24), pp.8788-8790. http://www.pnas.org/content/111/24/8788.full2