SLA Pitch

SLA Pitch What is fairness in machine learning? What kind of answer are we looking for to that question? Not an objective definition of a concept chosen from among competing alternatives. But rather what does it mean that there is a burgeoning subfield called fairness in machine learning? 

Academic fields are both communities of researchers and assemblies of ideas. So this is a social phenomenon, and it is a configuration of ideas. One way of saying what fairness in machine learning is is by looking at where its ideas come from. Positionality not just for people but for ideas. 

What I’d like to do

proposed study

@@ -280,22 +283,19 @@ class PlotModelComparison(DiversityInScienceTask):      “Plots model diagnostics with final model stats overlaid”      def run(self): -        diagnostics, model = self.input() -        model = LdaMalletPlus.load(model[‘model’].path) -        stats = pd.read_csv(model[‘stats’].path) +        diagnostics, modelData = self.input() +        model = LdaMalletPlus.load(modelData[‘model’].path) +        modelStats = pd.read_csv(modelData[‘stats’].path)          ddf = pd.read_csv(diagnostics[‘df’].path)          f, (ax0, ax1) = plt.subplots(1,2, figsize=(10,4))          ddf.groupby(’num_topics’).mean().coherence.plot(ax=ax0)          (ddf.groupby(’num_topics’).mean().exclusivity * ddf.num_topics.unique()).plot(ax=ax1) -        ddf.groupby(’num_topics’).mean().log_likelihood.plot(ax=ax2)          ax0.set_ylabel(“coherence”)          ax1.set_ylabel(“exclusivity”) -        ax2.set_ylabel(“log_likelihood”)          ax0.scatter([70], [modelStats.coherence.mean()], color=‘orange’)          ax1.scatter([70], [(modelStats.exclusivity.mean() * 70)], color=‘orange’) -        ax2.scatter([70], [modelStats.coherence.mean()], color=‘orange’) -        plt.savefig(self.output()[‘chart’].path) +        plt.savefig(self.output().path)