This paper aims to perform a portfolio optimization under the Mean-Semivariance Behavioral Hypothesis and measures whether there is dependence between the optimal portfolio structures thus obtained and ESG Risk Scores. The investigation of such an issue may be justified by the fact that ESG reporting is intended to become a stable evolutionary strategy (in the sense of Smith & Price), and portfolio optimization under such a behavioral hypothesis is in line with behavioral finances, where investors are considered to be twice as sensitive to losses as to gains (in the sense of Kahneman & Tversky). Following such an approach expressed methodologically and empirically, the result we reach, on the data we analyzed, is that: optimal portfolio structures are dependent on ESG Risk Scores and even if this statistical dependence is considered to be of low intensity we observe a pattern. The dependence is in the opposite direction up to the portfolio tangent to the Sharpe Efficient Frontier (the portfolio with the maximum Sortino ratio), after which, the dependence is in the same direction. We also observe other patterns in the analysis.