Glad to hear you like the idea. My method for reading through HN and others is to scan through items with a certain threshold. Once I get below the "fold" of the page, I start to lower the level so I can compensate for the "decay factor" you mention.
There's tons of good gems at the bottom of the page, and they need to compete at the same level as those at the top.
You could tweek the "weighting equation" and try linear, quadratic, or exponential type formulas.
You could also apply this technique to articles, so those at the top don't get overly hyped and those at the bottom have a fair chance to making some movement to the top.
Also, one thought in terms of the scoring value is to have it with a few digits of decimal precision. So you could still see how competing articles are doing at the very top and see your vote make a difference (be it very small).
