Experimental algorithmic modifications across a three-month period during the 2020 US election produced no detectable changes in participants' political attitudes