The gifts delivered and the rifts created by personalization
Sometimes countervailing arguments on a singular topic could stump us while being interesting. It becomes interesting when that can be linked to a common theme.
The contrasting events…
Late last month I had the opportunity to present a session on AI/ML and its implication for business strategy to a cohort of mid-career professionals at a leading management institute. It was a participative three-hour session. The session was peppered with case studies covering a couple of industries where the deployment of AI/ML had created a significant business impact.
The section of the session that was engaging was the case study on fashion e-tailing, (Stitch-Fix.com). Arguably, it was the high point of the session. The spotlight was on the topic of “Mass Personalization”. The data-led, AI/ML-driven strategy based on user and usage, enables the calibrated and customized reach providing the individual appeal to the multitude of consumers, of the e-tailer. Consumer acquisition, inventory management, logistics, and reverse logistics, addressing the temporal needs of the user, are executed with the smart usage of the made-for-purpose AI/ML algorithms. The level of sophistication delivering the personalization at the scale of operations makes it a compelling story.
Will come back to this point shortly.
The ubiquitous smartphones and the social media-inspired messaging are made for each other. If one is on social media being in a group or two will not be far away. Most often the Fear of Missing Out (FOMO) provides that gentle nudge to get enlisted in the alumni or community groups. Once in the group always in it. One can sneak out of a party unnoticed but not so from a social media group, without creating a flutter!
In the social media platforms, the group essentially comprises four categories of participants. A small fraction would influence and drive the discourse, the activists (the prime movers). Often, the narratives would be compelling. They could veer around politics or religion and sides would be taken — “North” and “South”, “Right” and “Left”, “Progressives” and “Conservatives”. At times, it is amazing to see the litany of inputs shared in real-time to buttress each other’s views and counterviews. Agitation and coagulating into groups make the next phase. It would be difficult to separate the truth from fiction. The discourse now noisy, would be egged on by the supporters (the patrons) mostly through emojis. The rest would be passive observers (the non-vocals) who could be ignoring the discussion, grudgingly accepting, or welcoming the diversity of views. The last category (the referee)- the “active” observers, would only intervene, should matters become toxic. A joke, a quote, or a comic strip would be thrown into the mix to douse the “fire”. A potential polarization is averted.
I was intrigued as to how the inputs could be brought about on the fly to fuel, provoke and sustain the discussion.
The voyeur plays God!
This led me to the topic of “Filter Bubble” (the phrase famously and formally propounded by Eli Parser through a TED talk). Here, he presents a persuasive argument on how algorithms unilaterally decide what we “want” to see based on our browsing habits, rather than what we “need” to see to get a balanced view. In a short sentence, it is the algorithm, which could suppress or amplify certain conversations. The choices are being invisibly manipulated, resulting in a confirmation bias.
The voyeur plays the God!
This has an unintended consequence. A self-reinforcing spiral, resulting in a unique online universe for the individual — the Filter Bubble as I understood it.
The enabler here is the algorithms that generate the unique content suggestion through the recommender systems with the personalized user interface. This invariably leads to a personalized view of things we “want “to see.
Filter Bubble does have an outsized at times deleterious effect on political, social, racial, and religious leanings. It can potentially create a rift between societies and people, to the point of fraying the fabric of the community. This is what we could see in the messaging groups, I referred to earlier. This may make it look like all is wrong with the algorithmic bias leading to confirmation bias.
A non-sequitur fallacy?
Courtesy: Fusioninformatics.com
Giving away to Get?
Coming back to the background preparation for my lecture session. The recommendation engine (for the content) with its personalized output (for the user) was a great facilitator in my preparation. Similarly, at a larger scale, the example of the towering success of the e-tailing fashion company, Stitch-Fix is attributed to the excellent execution of “mass personalization”. Could this personalized storefront for millions of users be created without the algorithm-led recommendation engines? A master stroke in the notoriously fickle world of e-tailing bringing in the much-valued “consumer stickiness”.
Let us for a moment dwell on the benefits derived, through this well-targeted campaign, filtering out irrelevance and bringing in “narrow self-interest”. Enabled by the algorithm, this gets the user what she wants and not necessarily what she may need.
Internet users, who use the internet for activities like buying online or doing research on focused areas are indeed benefited from the recommendation engines. The context and collaborative filtering provide benefits, which possibly get overshadowed. We may not be acknowledging it enough.
If one wants to buy an expensive sandal at the desired price one could browse the e-tailer’s website. This “narrow interest” of the prospective customer is addressed with personalized emails from the prospective e-tailer and with ad pop-ups (annoying targeted advertising). The browser is alerted with options when the offer is attractive or when new models are launched. This targeted advertising gets a steal of a deal. Small inconveniences and annoyances if tolerated, could yield a gift of sorts.
Context Matters!
So, Filter Bubble and the resultant personalization could be providing a Gift to someone but could be the cause of a Rift in some other context. The context decides that.
Privacy is prized. Are we willing to give away a sliver of it for the gift of personalization? Technology is an enabler. Using human judgment could help us employ it to our advantage.
We can have the cake and eat it too!
___________________________________________________________________
Somjit Amrit
Email: somjitamrit@gmail.com
