Self-report è Meglio delle Misure Implicite?

“Indirect behavioral measures in which participants do not have to introspect and report on their thoughts and feelings”

“Today, there are more than two dozen implicit measures, including the implicit association test, the affect misattribution procedure, and the evaluative priming task. These measures have been used to assess a variety of constructs such as attitudes, stereotypes, self-esteem, motives, outcome expectations, values, self-harm ideation, and semantic relations between concepts”

“First, mental contents are inferred from performance indicators (for example, response times, error rates or response frequencies) typically measured via computers. Second, responses on implicit measures are assumed to be driven by automatic processes (processes that are unintentional, unconscious, efficient, or uncontrolled)”

• No contamination (desiderabilità sociale, influenza del contesto)

• No consciousness (accesso a pensieri/sentimenti che sfuggono all’introspezione)

• Automaticity (accesso diretto a processi mentali automatici)

• Simple associations (es. proposizionalmente so che l’associazione tra “vaccini” e “autismo” non esiste nella realtà, ma nella mia rete semantica i due concetti potrebbero essere associati)

• Robustness (accesso a processi di apprendimento associativo molto radicati, stabili nel tempo)

• Person vs. environment - teorizzazioni in realtà contrapposte (misurano caratteristiche individuali o dell’ambiente [es. stereotipi comuni]?)

• Desiderabilità sociale in self-report è un problema ma 1) la rilevanza è inflazionata (totalità anonimità può anche peggiorare qualità ricerca); 2) il problema esiste anche nell’implicito (es. effetti cambiano in base a gruppo/minoranza sperimentatore)

“one study found that public (versus private) testing conditions reduced scores indicative of anti-gay prejudice on a gay/straight implicit association test to a larger extent (d = .40) than self-reported homophobia (d = .17) and heterosexism (d = .35)”

• Inoltre, “implicit measures are notoriously sensitive to faking strategies. Social biases on implicit measures are lower when participants are told that their “job is to be as nonprejudiced as possible” compared to when they are not”

• Infine, alti punteggi di desiderabilità sociale non sembrano portare a minori correlazioni implicito-esplicito.

generalmente scarsa nelle misure implicite, “the implicit association test and the affect misattribution procedure are the only two instruments that show estimates of internal consistency that meet standard psychometric criteria with Cronbach’s alphas between .70 and .90. All other instruments suffer from substantial measurement error with modal Cronbach’s alphas around .50.”

“Moreover, even the two implicit measures that have shown acceptable estimates of internal consistency (the implicit association test and the affect misattribution procedure) show lower test-retest stability (rs = ~.50) than conceptually corresponding self-report measures (rs = ~.75)”

Le correlazioni tra “single feeling integer” e outcome a distanza di un anno variano in base ad outcome e contesti, ma in media e mediana sono 0.15 e arrivano a 0.20, superando generalmente le variabili socioeconomiche

• Risparmio di tempo, “why one should administer the same implicit measure multiple times when the same information can be obtained with a simple self-report measure—especially when self-report measures based on a single item can reasonably predict (r = ~.15) important life events (for example, job change or relationship dissolution) one year ahead of time” [cf. slide precedente]

• Self-report non ha bisogno di basarsi su categorie contrapposte

• Self-report può essere usato per concetti complessi (es. atteggiamento verso a proposed policy change) non facilmente rappresentabili in task impliciti

• Self-report non richiede strumenti tecnologici e software