Executive summary: I believe the low scores may be because there is a shift to give the 0-10 scale proper meaning so that 5/10 means "average game" rather than 8/10 meaning "average game".
Now, on to the wall-of-text.
Funny to run across this article today. My friend and I, just last night, were discussing game review ratings inflation over a meal at Red Robin. We both came to the consensus that if the 0-10 scale is being properly used, then most games, at any given point in the history of the game industry, should fall in the mid 4 to high 6 range of scores. The reason is that this range, no matter the context, criteria, and time period, should always correspond to an average game. This is a game that hits all the basics, doesn't really do anything wrong, but doesn't do anything spectacular or revolutionary either. These are games that simply make par. Once you get into the range of 7's, you start talking about good games that maybe do some interesting stuff. At 8, the games become extraordinary works that have a high level of quality and polish. At the range of 9 and beyond, these are games that are works of genius and pure masterpieces. These would be the kind of games that reinvent the genre or completely change the rules of how games of that type (or, sometimes, any game) are designed and made. These are truly wonder works.
Now here's the problem. Once the concepts, techniques, and ideas of the 8, 9, and even 10 rating games becomes common or expected, any game thereafter can only garner a score, once again, in the mid 4 to high 6 range, possibly even low to mid 7's if enough polish is put into it. This is because the spectacular becomes ordinary (or mediocre) once it becomes common enough and easily reproducible through minimal effort. It becomes the norm. Consider the fact that many of the elements in games today that we take for granted, just 10-15 years ago required incredible effort and truly extraordinary, out-of-the-box creative thinking to accomplish. Nowadays, much of the effort to produce those same elements is a matter of calling some function built-in to the hardware or the packaged game-engine. It's a mundane effort to create those same elements now. It just doesn't astound us or fill us with awe and wonder anymore. It's expected and ordinary; thus, it gets a score hovering in the mediocre range of mid 4s to mid 6s, maybe low to mid 7s, depending on the amount of extra polish put into the game.
At the heart of it all is a disconnect with reality. We're expecting games to be mind-blowing at every single iteration, and that's just not possible. We are constantly needing a bigger and bigger "rush" for each successive game to satisfy us (it has to be more potent each time). Unfortunately, game quality, like so other things, likely follows a Bell curve. Most games, no matter the context or time period, are going to fall into the average range, scores of 4.5 to 6.5. As you go outside that range, in both directions of decreasing quality and increasing quality, you get a rapid, monotonic decrease in the population of games at those scores. If we are always expecting most games to be well-beyond average in quality, i.e. beyond what is considered the norm, then we are only setting ourselves up for disappointment each time.
Even further to this point, we also need to consider that much has been made of the visible score inflation that has been going on in years prior, such that an 8/10 score is considered "average". This kind of skewing, once corrected, will make it appear as if games are of lower quality than prior years if we are still holding to the notion that 8/10 is for an average game, as opposed to a more proper 5/10 rating. The games are likely no better or worse, in context, than at any other time; however, our perceptions have been distorted by long-standing abuse of the scoring scale to give the appearance that every game is of exceptional quality, i.e. worthy of 8/10, when, in fact, the game is only of contextually average quality. This is what leads to the phenomenon of "8 out of 10" becoming "Hate out of 10", as Jim Sterling coined the term. We can see that the games are only of contextually average quality, so we mentally equate 8/10 to "average" or mediocre, even though the scale says that 8/10 is supposed to be exceptional. We still try to use 8/10 as meaning exceptional, but we know this is not really the case by observation of the population distribution. Thus, the meaning of the scale has been disconnected from reality.
I would posit that the phenomenon we are seeing here with these scores is a long overdue correction in which the meaning of the 0-10 is properly aligned with reality such the 4.5-6.5 range means "average", in context, and scores of 8.0 and above are the truly exceptional games. Because the population distribution can be expected to follow something similar to a Bell curve, we should see the vast majority of games occupy scores in the 4.5-6.5 range, with progressively fewer games scoring outside that range, higher or lower. This could make the appearance of games being worse, if we hold to our prior distortion of 8/10 being an average score, when they are, in fact, contextually unchanged in quality.
EDIT: Decided saying "For the TL;DR" sounds to insulting. Changed it to be what it really is, an executive summary.
ADDENDUM: I have commented in prior posts my disagreement with the level of precision that reviewers attempt to assign to game-review scores. In my opinion, a 1 in 100 or better precision is just not possible with a qualitative heuristic like reviewing a game. There simply is no scientific standard criteria or mathematical algorithm, in my opinion, being used universally for reviewing games that could allow such a high precision in the scoring process, even with averaging of multiple scores from multiple reviewers. At best, one can only meaningfully attain a 1 in 10 precision, and even degrading to a 1 in 5 precision is often sufficient to convey qualitative meaning to the score. In my opinion, the qualitative nature of these scores just doesn't lend credence to using such high precision in the numerical value.
