Several glaciers that had previously been buttressed by the Larsen B ice shelf accelerated by a factor eight after said ice shelf disintegrated. The demise of this ice shelf lead to an increase of 27km3 of ice loss per year (Rignot et al., GRL 2004).

This is based on a combination of observations and modelling (see e.g. Joughin et al., Science 2014). Observations show us several glaciers are in decline (e.g. Rignot et al., GRL 2014) and ice shelves, which serve as a buttress for the glaciers, are thinning due to warmer sea water. This process is seen to be accelerating (e.g. Paolo et al., Science 2105). Where ice shelves (like Larsen B) have already vanished, glaciers in the hinterland have indeed sped up (Scambos et al., GRL 2004). Numerical models predict that the changes underway now are likely to lead to a full-scale collapse of the west Antarctic Ice Sheet.

Six feet by 2100 is a very high estimate; the earlier mentioned study by DeConto and Pollard doesn't go above 1.5 m under 936 ppm atmospheric CO2. At 538 ppm their estimates don't go above 86 cm by 2100.

A NOAA-produced graph of enl Nino intensity illustrates this:

The usual method of prying apart the various factors responsible for temperature changes is by modelling; we know the input of the sun, we know the composition of the atmosphere, etc etc. If you have a climate model that, with these parameters, can reproduce the climate of the past, you can be fairly sure that, even being by definition a simplification, it gets the physics right. You can then use the model to calculate what the influence of each individual parameter was (e.g. Huber and Knutti 2012, Anthropogenic and natural warming inferred from changes in Earth’s energy balance, Nature Geoscience 5, 31-36). You can even switch parameters on and off to see what happens. These models can then also be used for climate predictions under various assumptions on, say, GHG emissions.