Saturday, November 16, 2013

The Pitfalls of Mental Outsourcing

This post builds somewhat off my last thought. I introduced the idea of coherence generating mechanisms. These are tools meant to synthesize and understand some underlying pattern in the context of social phenomenon. In the realm of music, you can think about Pandora. Via a feedback system, it seeks to hone directly into ones preferences. The more I like R&B songs, the more the system understands that I have a preference for them. The brain also fits this definition. It’s one big pattern recognizer, trying its best to turn thought intensive responses into reflexive impulses. With the rise of “big data” these systems are becoming more ubiquitous, whether it be Google Ads or Netflix’s recommendation systems.

I think the best way to define something is to define what it is not. My idea of coherence generating mechanisms are not simple single lag systems. As in, they do not assume that the future will exactly mirror the past. Thus, the system has to “learn” in some sense. It must seek to create categories of knowledge that it calls upon to utilize in the future. Netflix’s recommendation system fits this since it attempts to utilize both viewing habits and ratings to pin point some locus of movie preference. Other examples will be discussed at length below.

From the standpoint of marketers, the utility of these mechanisms are clear. Strip away all the excess of human activity, and drill down to some innate dispositions and preferences. While I think it is clear people have preferences, be it liking Indian food relative Sischuanese cuisine, I think we need to acknowledge limitations and tradeoffs associated with these approaches. Think about listening to Pandora while you are running. This perfect song comes on that has a beat to keep you moving. (FYI any beats above 145 BPM seem to yield little effect). So you like that song on Pandora. Now three hours later, while sitting in your bedroom, contemplating the reason why cars park in driveways but drive in parkways, that same song comes on. Now in your contemplative state, that song comes off as too much, the beat pulses through your brain disrupting your ruminative state of mind. This situation can be easily resolved if you condition preferences on particular states of mind. But Pandora does not know what you’re currently in the mood for, it only knows the feedback you have generated for it. While these systems feel intelligent, they are simply classic stimulus-response mechanisms.

This feeling of something deeper going on in these systems gives rise to an even more important issue, how does the user feel about this system? What sort of confidence does the user place in it? These systems are built to allow individuals to outsource that impossible task of introspection. Rotten Tomatoes creates a single, unified metric for the quality of the movie. But under the guise of arithmetic precision hides all the underlying biases of the culture industry. Those movies that can garner reviews are the ones that end up on the site. Furthermore, band-wagoning and other group level effects, such as gender and racial bias, can skirt by under the false sense of security that numbers provide. Further studies on how users come to appreciate choices that are derived from these metrics are important, on top of, studies looking at how users utilize these informational channels in further searching behavior.

That the underlying metrics may be flawed does not distract from the way these systems have the potential of opening up users to new domains they have yet to experience. Google Maps gives me the confidence to walk to places I have never walked to. These systems open up the landscape for one to explore. From the perspective of learning, however, do these systems allow the user to internalize the new environment that they are exploring? Research into GPS systems impacts on mental visualizations shows that is not the case. GPS technology trades off with specialized mental tasks that underlie some important parts of cognitive development. Like exercising a muscle, the less users utilize their mental mapping, the harder it is to get it back in gear (Here is a short article detailing ways around this problem.)

These systems do not just have an ability to trade off with internal resources we use, but also resources in the external world, as well. Social communication used to be the route for recommendations in the past. Word of mouth was the primary tool to resolve informational asymmetries in the experiential good market. While studies have found that critic’s reviews seem to be driving revenue generation for films, the question of what these systems do to social ties are beyond the question of simple market analysis. Do we feel less of a need to consult one another because we have Rotten Tomatoes at our finger tips? Do we view our friend’s recommendations with more skepticism if they contradict the “popular” metrics we utilize online? If your friend constantly recommends CDs that MetaCritic deems to be terrible, does your friendship suffer? 

These are just a small sample of possible negative implications of outsourcing our own internal pattern seeking, or coherence generating, mechanisms. The ultimate question from an economic science perspective is do the benefits outweigh the costs. While I believe at the end of the day that most do, the lack of attention to some of the negative ramifications of these systems should cause pause when attempting to run your own cost-benefit analysis.  

Monday, November 11, 2013

Reflektions

I decided to try and squelch some of my more ADD tendencies, so I shut off all outside distraction, turned the lights down, and got to listening to the new Arcade Fire album while perched on my bed. Letting my mind wander, I started to pick up subtle nuances as the album developed. The melody slowly rising and overtaking the struggling Wynn Butler who seems early on in the album to be at war with the instrumentals, attempting to make his voice have meaning over the cacophony of sound. I thought to myself the lay out of this album is masterful, the transitions work beautifully, it builds from start to finish, just to come all crashing down as the first side ends. But this post is not about my feelings persay, it’s about the sense of coherence one feels when listening to an album in its entirety (which I will admit I didn’t even do since I stopped at the first side, my ADD sometimes always win out.)

Is the feeling of coherence intrinsic to the album itself? Most musicians and those around music would say definitely. The songs are placed in their order for some thematic reason that the authors have some privileged access to. It could be stylistic as in building up tempo, or it could be thematic as in many of the Killers’ early albums, taking you through a journey of pain and loss, ending on a bitter and exhausted note. What if a band such as Arcade Fire were to generate 40 songs and then dump them on their label who then have the job of ordering and building up a theme? Would the coherence one may feel from this be less valid than the coherence one feels from an album constructed coherently by the author? Do creation and construction live in separate realms?

To understand this question, one must look from where coherence comes from. For all the best laid plans a band has, people are fickle with regards to feeling how you want them to. The coherence of theme may be lost for the coherence of melody and tempo. In fact, single songs are sometimes put into coherent narratives by our own personal coherence generating mechanism, our brains. Many people experience the feeling of having their Ipod on shuffle, or listening to the radio, and that perfect song coming on at just the right moment. The song order was fundamentally random with no underlying structure, yet you the observer felt some deeper meaning in the songs arrival in your personal arc. In fact, many people hope to discover new connections in their music repertoire with external coherence generating mechanisms, such as ITunes Genius and Pandora. These systems attempt to synthesize an underlying coherence of the listener’s music tastes and map them onto either new or already existing songs within the listener’s repertoire.

Now it seems that I have come to divide creation and construction as two separate realms, in that construction can be reconstruction by the observers, the consumers of music. But this same reconstructing of narratives that listeners do, is the same sort of reconstruction bands do as they are creating albums, and songs. The tempo, notes, and beats of stanzas are constantly being morphed, blended, discarded, folded into larger themes, melodies, cadences of songs, which are then recombined, spliced, manipulated, lengthened into full-fledged albums with their own arcs, narratives and meta-structures. But is this process linear, does it build from stanza, to song to album without any feedback between these interlocking steps?

Music occupies a particular human world of creative expression. Though it is unique, it falls within the realm of human cognition and I think the answer to this quandary may lie there. A simple analogy to the process of language production may be illuminating. Whenever one blurts out something to their friend about how grand their day was, what comes first the meaning of the message or the content used to convey said meaning? Many people instinctively say that meaning comes first, for without the meaning, why would we express that particular phrase? But if the meaning came first, what language was the meaning in? If it was in English, then isn’t the meaning already in words, so where did the words come from to give to the meaning. Many researchers have come down on the side of the co-relationship between meaning and content. Both simultaneously constrain one another. 

The vocabulary we have at our disposal affects the meanings that we express and vice versa. There’s a reason people who speak more than one language often talk about thinking, or reasoning in another language so as to resolve some problem. Linguistic constraints beget reasoning constraints (Dennett, 247). Oliver Sachs has a wonderful, reporting of the constraining aspects of linguistic mediums:

“Communication by motor behavior became a very important part of the transference…[W]ithout knowing it, I was receiving two sets of communication simultaneously: one in words, a form in which the patient ordinarily communicated with me; the other in gestures [signs], as the patient used to communicate with his father. At other times in the transference, the motor symbols represented a gloss upon the verbal text the patient was communicating. These motor symbols contained additional material which either augmented or more likely contradicted what was being communicated verbally. In a sense, “unconscious material” was making its appearance in consciousness by way of motor rather by way of verbal communication (Sachs, 34).”

Transference is a bit much for me, but it is clear that the mode with which reasoning is articulated and rehearsed affects the thoughts that are produced. Verbal and motor communicative structures are different and can construct thoughts in ways that are sometimes at odds. The abacus and calculator are two different methods for the same goal “calculation”, yet the mechanisms used and results generated can diverge.

This same sort of feedback between meaning and content can be seen in the relationships between albums, songs, and their listeners. While the artist may be developing the album, the album is also being particularized in songs, and once those songs are developed they themselves determine where the album is going. This is why outlining is such an important part in the writing process. For all the jumbles in one’s heads, an outline helps to particularize and focus those disparate connections into some sequential process. Once the album is created, the meaning of the artist is now supplanted by the meaning of the listener. The listener brings in their own personal narrative, but those are themselves constrained by the content of the album. There is a reason people do not usually feel particularly sad listening to a Katy Perry album, the content to derive that meaning is not there. The construction/creation divide (or more generally the form/content divide) is less a divide and more of an ecosystem of production that highlights the multilayered process of reasoning. 

-----------------------------------------------------------
References:

Dennett, D. Unconcsiousness Explained. Little, Brown and Company: Boston. 1991.

Sachs, O. Seeing Voices: A Journey into the World of the Deaf. Harper Collins: New York. 1990.