Modelings decisions in AI:re-thinkning Linda in terms of coherent lower and upper conditional previsions