Sunday Mumble

And it's Aug 08 today. I don't know when it started to become a tradition, but various popular online shopping platforms in SIN city have created monthly ``special sales'' events surrounding dates which have the same month and day. And their marketing blitz has always been relentless. I remember that when I was still working at my old place, the advertising screen at the lift lobby would always be playing some heinous version of one of the shopping platform's jingle, with the almost customary mid-riff revealing background dancers who look like they are high on something even as the entire jingle was basically the shopping platform's name repeated ad nauseum.

Thankfully, I don't get to see such ads when moving about online from my phone nor PC. My beef with advertising over online services comes from these few perspectives:

1. Unlike television, each time these advertisements play, they consume bandwidth that I paid for, which is made worse when the bit-rate quality of these advertisements are higher than the source media that I am looking at;
2. These advertisements tend to be obnoxious in their bid to gain attention, from being generally louder than the content that is being played, to taking up the whole bloody screen and locking controls on my web browser/screen;
3. There relevance has always been missing, and are generally tone-deaf in nature; and
4. The laziness of some content creators/platform owners in curating these advertisements, relying on surveillance and the black-box of data mining/machine learning to feed third party advertising networks that have no interest in doing good, leading to yet another attack surface for malware.

I mean, if a company is sponsoring a content creator (whether video maker, live streamer, or writer), and the content creator gives them a shout-out from within their video, I will generally sit through it to learn about their product/service, for the two reasons of (a) they are supporting the content creator, which is relevant to me; and (b) the content creator trusts the company's product/service enough to plug for them, which again raises the relevance factor to me.

I think the whole premise of ``we must collect all data from everyone so that we can know their preferences and determine relevance'' is incorrect to begin with. It makes the assumption that the consumer is a dumb-ass who does not know what he/she wants, and therefore needs the use of fancy statistics to ``inform'' them what it is that they are missing. It takes agency away from the consumer and forces a top-down authoritarian approach towards choices.

``Corporations ain't God, and so I don't trust their moral compass,'' I say as I write and publish my blog entry on one of the platforms that is owned by a corporation.

I think that this clash between ruling purely from statistics against being guided by the statistics while acknowledging individual choice is that drives much of the recent debates/non-debates, particularly on pandemic-related instructions. Statistics always describe the large-scale population-level behaviour, but when it comes down to the individual level, it will collapse to an actual outcome that can be colloquially called ``defying statistics''.

Classic example: the total fertility rate (TFR) of 0.84 in South Korea in 2020. This means the average number of children that would be born to a woman over her life-time (subjected to various assumptions that are unnecessary for my point) in South Korea is 0.84.

But what is 0.84 of a child? There is no such thing. Each woman can only have a non-negative integer child over her life-time---the statistic of TFR does not correspond to any single outcome for each woman. If we can understand such a concept for something like TFR, why is it so hard to realise the limits of the statistical black-box that is machine learning/data mining and not oversell the bloody thing too much?

``Okay MT, so assume I get your point. Using population-level statistics to rule individual actions directly is probably a bad idea---let's say I agree with that. Isn't that why they are trying to improve upon that by gathering more data per person so that they can figure out the details to guide the person to the right outcome then?''

That's assuming that there is some consensus on what the ``right'' outcome means. For public policy, having society turning towards a certain direction that increases the maximal amount of happiness/contentment/some other society-agreed-upon criteria might be the ``right'' outcome to justify the level of data collection [per person]. Things get much murkier when we get to the corporation level. To them, increasing profit through either increasing revenue or reducing expenditure is the ``right'' outcome for every decision that they make, ergo the people whose data are being collected are not part of the beneficiaries of the invasion into their personal space---in that sense, perhaps this ``right'' outcome for the corporation isn't really the ``right'' outcome for the people then.

The arbiter of that exists, but unfortunately He operates at a scale that is literally beyond our ken, so we are stuck with working together for now. This means that such data-driven ``guidance'' will always be a tightrope walking exercise between what the numbers say against how the people will feel about the outcome with respect to the quality of life. I think that at the very least, we can all agree that striving for a positive quality of life is usually the default goal, with the main disagreement on what time scales we are looking at, and what constitutes as being a ``positive'' quality of life, both of which are contextually determined by the socio-economico-politico environment, which in turn has deep roots in geography and culture.

I maintain that life is emergent and not stochastic nor deterministic, and the profound consequence of this is that a purely qualitative approch or a purely quantitative approach will never be sufficient to explain/guide the future, which implies that any system based on either of these two approaches alone is bound to fail.

The gamble then is to ensure that it fails after the time span of interest and outside of the space where accuracy is cared about. So far, I think they are losing this fight.

Will they ever win? I'm not going to prognosticate.

Tomorrow's a public holiday, and as one who isn't working now, it'll just be like any other day. That is all I have for now, and so, till the next update then.