2018-02-03 Thought stream
These notes are rough and partial, made during stops on a weekend stroll today. In facing the tradeoff of either sharing them now or waiting til I might come back and polish them, I opted to share now to engender dialogue with those in the grip of similar interests ---about perception, reading, voice applications, and cognition.
When I was young, I enjoyed few things more than to immerse myself in reading. But now I grow restless after a paragraph or two of academic text, even though I'm eager to absorb the whole book. A few things have changed: I enjoy green tranquility more than I did then, and I try to immerse myself in it every day. My lenses have grown less flexible and I'll get reading glasses soon.
I've always been interested in perception, and how people with sensory disabilities invent “life hacks”. All this has propelled me into exploring alternate modes of reading.
To give my eyes a rest, I commute by bus Berkeley<=>SF and get spectacular views from the Bay Bridge
To further enrich this time, I've searched for audiobooks. Highly recommended: Ka by John Crowley; Louder than Words by Ben Bergen.
I'm still in search of adequate mobile private audio. IQBuds fall out of my ears too easily. Shark putty on in-ear buds works okay but a friend tells me it may lead to hearing loss. Over-ear headset with ear hooks looks promising.
One frustration with earbuds is that Google Assistant doesn't work with them due to a long-standing bug. My day job involves voice user interfaces (VUIs) and I'd like to embed them more into my daily life to learn more about their potential (as I've blogged about before). The Alexa app might be an alternative but I don't want to give my personal data to both Google (calendar, keep) and Amazon (purchases, Alexa app).
I really need a way to block restaurant noise, and sometimes boost others speech at the same time. I had hoped HereOne (defunct) or IQBuds (didn't mask as much as I need) would be the solution.
Peter Vishton's video lectures on perception are a great complement to Louder Than Words, as is Baars' book In The Theater of Consciousness
The sense of touch trains vision about distance and texture
The phenomenon of parallax trains vision about occluded, unseen facets
I never understood until recently why some cognitive scientists thought there was a 'problem of consciousness'; they seemed to be asking 'how can we be in a state of being awake?'. There was no mystery in this question for many other cognitive scientists because, I believe, we all think this question is easily answered by analogy to a production system (of rules) where it becomes active if there are new premises/assertions/percepts and becomes quiescent otherwise (and fatigue forces us to shut off percepts regularly). But if they had said instead that many actions are driven by unconscious percepts rather than co-occurring conscious ones (e.g. reaching for an object involved in an Ebbinghaus illusion), I think we would have seen their point.
To aid my understanding of perception and communication, I'm formulating a set of rules/grammars to clarify what is being inferred from what, because as Vishton remarks, perception is inherently inferential. Some of the key micro-theories are: The physical world has a continuous series of energy radiations that emanate from some sources at particular distributions of intensity, frequency, and direction. These impinge on the human body which transduces sub-ranges of these energy distributions into sensory distributions; for example, we can sense a sub-range of light frequencies, and only in the direction our two retinas are pointed, and mostly via the foveas of those. Our body can 'unitize' some sub-ranges of these sensations into chunks, and some chunks into other chunks of wider coverage (including across sensors). A chunk persists in memory longer than a mere sensation (terminal chunk?). Some of the inferred chunks become available as percepts to consciousness.
Other chunks may be available only to drive bodily action. The abductive inference of these chunks continues past the boundary of consciousness as chunks organize our awareness of the actions of others, influenced by what we think were our intentions when we performed similar actions (or if we were to).
Although communication theories often focus on a communicating dyad, I believe this is too complex for an initial scenario and theory-forming should instead focus on a dyad where one of the persons is an unseen bystander. Then to a scenario where the bystander is detected by another unseen bystander. Then to a scenario where the observed person picks up clues that she is being observed. Then to the bystander realizing the observed's guess and hiding his attention before being caught. All seen by the 2nd bystander. Then to the scenario of mutual observation. Only at this point are we ready to realize the importance of 'active listening'. And only then (or just before?) are we ready to fully appreciate actions performed for the benefit of mutual observation.
A great walk in Berkeley for nature and architecture: (Install Flush app to find pit stops) Start at Berkeley Bowl on Oregon St and head down Russell. (Stop at Nabolom Bakery?) Turn left onto Piedmont. Turn left on Optometry Lane and walk past the Women's Faculty Club along the creek. Enter the Ohlone Greenway at Hearst Ave. (Stop at SweetGreen? Au Coquolet Cafe? Sat Farmer's Market?)
What I call 'chunks’ above (following their use by John Anderson’s cognitive modelling work) are called 'imago-schematic templates’ (or something like that, IIRC) in George Lakoff’s writings. Although Bergen doesn’t use either of these terms in his 'Louder Than Words’ audiobook, the book is largely about them. Bergen provides a tutorial in the theory of embodied cognition, which is opposed to symbol-oriented models like Anderson’s. But Being aware of chunks while hearing about imago-schemas made me wonder: How can a schema encode attributes, roles, and quantities? How are schemas indexed to aid retrieval? How can an unsatisfied part of one schema be fitted by part of another to enable abductive reasoning? Even though proponents of embodiment lean toward sub-symbolic connectionism as their preferred computational model, I think chunks are a better computational model. (Anderson’s model incorporates a sub-symbolic part also.)
The number of audiobooks worth listening to seems low; most titles I search for have no audiobook version. To do: See if Play Books or the Kindle app have adequate text-to-speech. Result: Android has a screenreader called TalkBack, and it’s compatible with PlayBooks and Kindle. But it won’t read more than a page at a time.
An idea: Chunks have different numeric functions that determine their activation level. For example, hearing one’s name gets a high activation over a short number of phonemes; hearing the initial words of a famous speech might “ring some bells” through many more phonemes at comparatively lower activations.
I think of chunking as an abductive inference process comparable to incrementally building a parse of a token stream. The 'head’ of such a parse/grammar rule is the precept/concept that we 'recognize’ as a consequence of matching (most of) the dependencies/antecedents/observables of the rule.
Two topics in Bergen’s book continue to occupy me: 1) the process of manipulating schemas for negative scenarios (e.g. “There is no gift on top of the fridge”) where there are measurable effects for the *positive* scenario for hundreds of milliseconds, and 2) how space can be used as a metaphor for time but not the reverse, even though both are omnipresent in our experiences that assemble into schemas.
To do: Review time perception, maybe in Stanford Encyclopedia of Philosophy (SEP). Vishton didn’t cover it.
