Behavioural_JJ science_NN1 :_: secret_JJ signals_NN2 You_PPY answer_VV0 the_AT telephone_NN1 and_CC hear_VV0 a_AT1 perky_JJ voice_NN1 :_: "_" Hello_UH ,_, this_DD1 is_VBZ Michelle_NP1 from_II Brown_NP1 's_GE computers_NN2 ._. 
I_PPIS1 wonder_VV0 if_CS you_PPY have_VH0 a_AT1 few_DA2 moments_NN2 to_TO talk_VVI about_II some_DD of_IO our_APPGE new_JJ products_NN2 ?_? "_" 
Sales_NN call_VV0 !_! 
Your_APPGE first_MD impulse_NN1 is_VBZ to_TO slam_VVI the_AT receiver_NN1 down_RP ._. 
But_CCB --_JJ well_NN1 ,_, at_RR21 least_RR22 the_AT woman_NN1 is_VBZ polite_JJ ._. 
And_CC knowledgeable_JJ ._. 
And_CC ..._... 
Fifteen_MC minutes_NNT2 later_RRR ,_, as_CSA you_PPY give_VV0 the_AT ever-cheerful_JJ Michelle_NP1 your_APPGE credit-card_NN1 number_NN1 to_TO purchase_VVI a_AT1 new_JJ Super_JJ Zippy_JJ hard_JJ drive_NN1 ,_, you_PPY tell_VV0 yourself_PPX1 what_DDQ a_RR31 good_RR32 deal_RR33 it_PPH1 is_VBZ ,_, and_CC how_RGQ much_DA1 you_PPY 've_VH0 been_VBN needing_VVG the_AT thing_NN1 ._. 
Or_CC do_VD0 you_PPY ?_? 
Computer_NN1 scientist_NN1 Alex_NP1 Pentland_NP1 from_II the_AT media_NN lab_NN1 at_II the_AT Massachusetts_NP1 Institute_NN1 of_IO Technology_NN1 (_( MIT_NP1 )_) in_II Cambridge_NP1 would_VM argue_VVI that_CST your_APPGE decision_NN1 had_VHD very_RG little_DA1 to_TO do_VDI with_IW what_DDQ the_AT saleswoman_NN1 said_VVD ,_, and_CC everything_PN1 to_TO do_VDI with_IW how_RRQ she_PPHS1 said_VVD it_PPH1 --_JJ her_APPGE intonation_NN1 ,_, say_VV0 ,_, or_CC the_AT fluctuating_JJ pace_NN1 and_CC amplitude_NN1 of_IO her_APPGE voice_NN1 ._. 
The_AT sale_NN1 was_VBDZ probably_RR made_VVN within_II the_AT first_MD few_DA2 seconds_NNT2 ,_, he_PPHS1 says_VVZ ,_, guided_VVN by_II signals_NN2 you_PPY did_VDD not_XX even_RR perceive_VVI ._. 
"_" Human_JJ behaviour_NN1 is_VBZ much_RR more_RGR predictable_JJ than_CSN is_VBZ generally_RR thought_VVN ,_, "_" says_VVZ Pentland_NP1 ._. 
A_AT1 person_NN1 's_GE responses_NN2 can_VM often_RR be_VBI explained_VVN by_II "_" non-linguistic_JJ behaviours_NN2 of_IO other_JJ people_NN and_CC simple_JJ instincts_NN2 for_IF social_JJ display_NN1 and_CC response_NN1 ,_, without_IW any_DD recourse_NN1 to_II conscious_JJ cognition_NN1 "_" ._. 
This_DD1 'second_NN1 channel'_NN1 of_IO human_JJ communication_NN1 acts_VVZ in_II parallel_NN1 with_IW that_DD1 based_VVN on_II rational_JJ thinking_NN1 and_CC verbal_JJ communication_NN1 ,_, and_CC it_PPH1 is_VBZ much_RR more_RGR important_JJ in_II human_JJ affairs_NN2 than_CSN most_DAT people_NN like_VV0 to_TO think_VVI ,_, says_VVZ Pentland_NP1 ._. 
Yet_RR evidence_NN1 for_IF it_PPH1 has_VHZ been_VBN accumulating_VVG in_II the_AT laboratory_NN1 for_IF decades_NNT2 ._. 
And_CC now_RT modern_JJ technology_NN1 has_VHZ made_VVN it_PPH1 possible_JJ to_TO monitor_VVI these_DD2 social_JJ signals_NN2 in_II spontaneous_JJ ,_, natural_JJ settings_NN2 on_II a_AT1 scale_NN1 that_CST was_VBDZ not_XX possible_JJ before_RT ._. 
"_" Collecting_VVG this_DD1 kind_NN1 of_IO spontaneous_JJ data_NN is_VBZ extremely_RR hard_JJ ,_, "_" says_VVZ neuroscientist_NN1 Anna_NP1 Esposito_NN1 of_IO the_AT International_JJ Institute_NN1 for_IF Advanced_JJ Scientific_JJ Studies_NN2 in_II Vietri_JJ sul_NN1 Mare_NN1 ,_, Italy_NP1 ._. 
"_" But_CCB Alex_NP1 has_VHZ found_VVN a_AT1 way_NN1 to_TO do_VDI it_PPH1 ._. "_" 
"_" Human_JJ behaviour_NN1 is_VBZ much_RR more_RGR predictable_JJ than_CSN is_VBZ generally_RR thought_VVN ._. "_" 
Alex_NP1 Pentland_NP1 Indeed_RR ,_, over_II the_AT past_JJ decade_NNT1 Pentland_NP1 and_CC his_APPGE team_NN1 have_VH0 developed_VVN a_AT1 range_NN1 of_IO wearable_JJ devices_NN2 that_CST include_VV0 electronic_JJ badges_NN2 ,_, modified_VVD personal_JJ digital_JJ assistants_NN2 (_( PDAs_NP2 )_) and_CC specially_RR configured_VVN mobile_JJ phones_NN2 ._. 
These_DD2 instruments_NN2 can_VM track_VVI not_XX only_RR a_AT1 person_NN1 's_GE location_NN1 ,_, but_CCB also_RR his_APPGE or_CC her_APPGE behavioural_JJ details_NN2 --_NN1 such_II21 as_II22 intonation_NN1 and_CC upper-body_JJ movements_NN2 ._. 
The_AT resulting_JJ data_NN can_VM then_RT be_VBI analysed_VVN for_IF patterns_NN2 of_IO meaningful_JJ social_JJ signals_NN2 ._. 
Pentland_NN1 and_CC his_APPGE colleagues_NN2 have_VH0 already_RR done_VDN multiple_JJ experiments_NN2 with_IW this_DD1 technology_NN1 at_II call_NN1 centres_NN2 and_CC elsewhere_RL ,_, and_CC are_VBR now_RT moving_VVG into_II sites_NN2 such_II21 as_II22 banks_NN2 and_CC hospitals_NN2 ,_, where_CS they_PPHS2 collect_VV0 data_NN on_II hundreds_NNO2 of_IO people_NN over_II many_DA2 months_NNT2 to_TO reveal_VVI the_AT hidden_JJ communications_NN2 in_II organizations_NN2 ._. 
The_AT potential_JJ reward_NN1 ,_, Pentland_NP1 says_VVZ ,_, is_VBZ a_AT1 richer_JJR ,_, more_RGR complete_JJ and_CC more_RGR objective_JJ view_NN1 of_IO human_JJ interaction_NN1 --_NN1 with_IW our_APPGE inherent_JJ bias_NN1 toward_II what_DDQ 's_VBZ conscious_JJ and_CC verbal_JJ taken_VVN out_II21 of_II22 the_AT equation_NN1 ._. 
Out_II21 of_II22 control_NN1 "_" It_PPH1 is_VBZ difficult_JJ for_IF people_NN to_TO accept_VVI "_" ,_, says_VVZ psychologist_NN1 John_NP1 Bargh_NN1 of_IO Yale_NP1 University_NN1 in_II New_NP1 Haven_NP1 ,_, Connecticut_NP1 ,_, a_AT1 pioneer_NN1 in_II the_AT study_NN1 of_IO subliminal_JJ influences_NN2 on_II human_JJ behaviour_NN1 ,_, but_CCB much_DA1 of_IO a_AT1 person_NN1 's_GE everyday_JJ life_NN1 "_" is_VBZ determined_VVN not_XX by_II their_APPGE conscious_JJ intentions_NN2 and_CC deliberate_JJ choices_NN2 ,_, but_CCB by_II mental_JJ processes_NN2 put_VVN into_II motion_NN1 by_II their_APPGE environment_NN1 "_" ._. 
"_" A_AT1 gazillion_NN1 experiments_NN2 show_VV0 that_CST I_PPIS1 can_VM flash_VVI something_PN1 at_II you_PPY so_RG fast_RR you_PPY do_VD0 n't_XX see_VVI it_PPH1 ,_, yet_RR the_AT information_NN1 does_VDZ bias_VVI you_PPY towards_II one_MC1 decision_NN1 as_II31 opposed_II32 to_II33 another_DD1 ,_, "_" says_VVZ cognitive_JJ psychologist_NN1 Michael_NP1 Gazzaniga_NP1 of_IO the_AT University_NN1 of_IO California_NP1 at_II Santa_NP1 Barbara_NP1 ._. 
Moreover_RR ,_, he_PPHS1 says_VVZ ,_, the_AT literature_NN1 is_VBZ full_JJ of_IO experiments_NN2 showing_VVG that_DD1 conscious_JJ explanations_NN2 for_IF our_APPGE behaviour_NN1 are_VBR often_RR just_JJ rationalizations_NN2 invented_VVN after_II the_AT fact_NN1 ._. 
He_PPHS1 cites_VVZ the_AT example_NN1 of_IO a_AT1 patient_NN1 whose_DDQGE corpus_NN1 callosum_NN1 had_VHD been_VBN severed_VVN as_II a_AT1 treatment_NN1 for_IF epilepsy_NN1 ,_, making_VVG it_PPH1 impossible_JJ for_IF one_MC1 side_NN1 of_IO the_AT brain_NN1 to_TO communicate_VVI with_IW the_AT other_JJ ._. 
Gazzaniga_NN1 and_CC his_APPGE colleagues_NN2 presented_VVD the_AT word_NN1 'walk'_NN1 to_II the_AT patient_NN1 's_GE left_JJ visual_JJ field_NN1 ,_, which_DDQ corresponds_VVZ to_II the_AT right_JJ side_NN1 of_IO the_AT brain_NN1 ._. 
When_CS the_AT patient_NN1 stood_VVD up_RP and_CC began_VVD walking_NN1 ,_, they_PPHS2 asked_VVD him_PPHO1 why_RRQ ;_; the_AT right_JJ side_NN1 of_IO the_AT patient_NN1 's_GE brain_NN1 had_VHD been_VBN shown_VVN to_TO lack_VVI the_AT ability_NN1 to_TO process_VVI language_NN1 ._. 
His_APPGE left_JJ brain_NN1 ,_, which_DDQ never_RR received_VVD the_AT walk_NN1 command_NN1 ,_, but_CCB which_DDQ handles_VVZ language_NN1 processing_NN1 ,_, quickly_RR invented_VVN a_AT1 logical_JJ explanation_NN1 :_: "_" I_PPIS1 wanted_VVD to_TO go_VVI get_VV0 a_AT1 Coke_NN1 ._. "_" 
Analogous_JJ effects_NN2 have_VH0 been_VBN found_VVN many_DA2 times_NNT2 in_II healthy_JJ people_NN ._. 
In_II the_AT early_JJ 1990s_MC2 ,_, for_REX21 example_REX22 ,_, psychologists_NN2 Robert_NP1 Rosenthal_NP1 of_IO the_AT University_NN1 of_IO California_NP1 ,_, Riverside_NP1 ,_, and_CC Nalini_JJ Ambady_NN1 of_IO Tufts_NN2 University_NN1 in_II Medford_NP1 ,_, Massachusetts_NP1 ,_, asked_VVD college_NN1 students_NN2 to_TO give_VVI reasons_NN2 for_IF liking_VVG or_CC disliking_VVG their_APPGE instructors_NN2 ._. 
The_AT students_NN2 tended_VVD to_TO mention_VVI the_AT instructors_NN2 '_GE friendliness_NN1 or_CC the_AT clarity_NN1 of_IO their_APPGE lectures_NN2 ,_, attributing_VVG their_APPGE rating_NN1 to_II conscious_JJ assessment_NN1 ._. 
But_CCB Rosenthal_NP1 and_CC Ambady_NP1 found_VVD that_CST most_DAT of_IO this_DD1 is_VBZ simply_RR rationalization_NN1 ._. 
The_AT researchers_NN2 could_VM predict_VVI how_RRQ around_RG 70%_NNU of_IO the_AT students_NN2 would_VM rate_VVI an_AT1 instructor_NN1 just_RR by_II analysing_VVG the_AT instructor_NN1 's_GE body_NN1 language_NN1 in_II 30_MC seconds_NNT2 of_IO soundless_JJ video_NN1 (_( N._NP1 Ambady_NP1 and_CC R._NP1 Rosenthal_NP1 ._. 
J._NP1 Pers_NP1 ._. 
Soc_NN1 ._. 
Psychol._NP1 64_MC ,_, 431-441_MCMC ;_; 1993_MC )_) ._. 
Pentland_NP1 's_GE efforts_NN2 to_TO track_VVI such_DA phenomena_NN2 in_II the_AT wild_JJ ,_, so_RR to_TO speak_VVI ,_, date_VV0 back_RP to_II the_AT early_JJ 2000s_MC2 ._. 
In_II one_MC1 experiment_NN1 ,_, the_AT group_NN1 took_VVD data_NN over_II a_AT1 period_NN1 of_IO several_DA2 months_NNT2 in_II 2005_MC ,_, as_CSA operators_NN2 at_II a_AT1 call_NN1 centre_NN1 in_II the_AT United_NP1 Kingdom_NP1 tried_VVD to_TO make_VVI sales_NN over_II the_AT telephone_NN1 ._. 
The_AT operators_NN2 were_VBDR fitted_VVN with_IW small_JJ electronic_JJ devices_NN2 that_CST measured_VVD variations_NN2 in_II their_APPGE tone_NN1 and_CC pitch_NN1 ,_, but_CCB not_XX the_AT specific_JJ words_NN2 they_PPHS2 used_VVD ._. 
Yet_RR the_AT researchers_NN2 were_VBDR able_JK to_TO devise_VVI an_AT1 algorithm_NN1 that_CST could_VM predict_VVI whether_CSW a_AT1 call_NN1 would_VM result_VVI in_II a_AT1 sale_NN1 from_II only_RR a_AT1 few_DA2 seconds_NNT2 of_IO data_NN ._. 
Successful_JJ operators_NN2 ,_, it_PPH1 turned_VVD out_RP ,_, spoke_VVD little_JJ and_CC listened_VVD more_RRR ._. 
And_CC when_CS they_PPHS2 did_VDD speak_VVI ,_, their_APPGE voices_NN2 fluctuated_VVD strongly_RR in_II amplitude_NN1 and_CC pitch_NN1 ,_, suggesting_VVG interest_NN1 and_CC responsiveness_NN1 ._. 
"_" If_CS operators_NN2 do_VD0 it_PPH1 right_RR ,_, "_" says_VVZ Pentland_NP1 ,_, "_" they_PPHS2 are_VBR almost_RR certain_JJ to_TO be_VBI successful_JJ ._. "_" 
Indeed_RR ,_, the_AT call_NN1 centre_NN1 now_RT uses_VVZ this_DD1 new_JJ understanding_NN1 to_TO train_VVI its_APPGE operators_NN2 to_TO sell_VVI more_RGR effectively_RR ,_, and_CC to_TO recruit_VVI people_NN who_PNQS have_VH0 the_AT right_JJ speech_NN1 patterns_NN2 ._. 
Other_JJ experiments_NN2 suggest_VV0 that_CST such_DA measures_NN2 could_VM improve_VVI sales_NN success_NN1 by_II as_CSA much_DA1 as_CSA 20%_NNU ,_, says_VVZ Pentland_NP1 ,_, which_DDQ is_VBZ worth_II millions_NNO2 to_II large_JJ industries_NN2 ._. 
And_CC in_II an_AT1 experiment_NN1 involving_VVG a_AT1 45-minute_JJ mock_JJ salary_NN1 negotiation_NN1 between_II students_NN2 in_II a_AT1 business_NN1 school_NN1 ,_, Pentland_NP1 says_VVZ that_CST by_II combining_VVG several_DA2 display_VV0 signals_NN2 from_II the_AT first_MD 5_MC minutes_NNT2 of_IO the_AT negotiation_NN1 ,_, his_APPGE team_NN1 could_VM predict_VVI who_PNQS would_VM come_VVI out_RP on_II top_NN1 with_IW 87%_NNU accuracy_NN1 ._. 
As_II a_AT1 result_NN1 of_IO such_DA experiments_NN2 ,_, the_AT MIT_NN1 group_NN1 has_VHZ identified_VVN a_AT1 handful_NN1 of_IO common_JJ social_JJ signals_NN2 that_CST predict_VV0 the_AT outcomes_NN2 of_IO sales_NN pitches_NN2 ,_, the_AT success_NN1 of_IO bluffing_VVG in_II poker_NN1 ,_, even_RR subjective_JJ judgements_NN2 of_IO trust_NN1 ._. 
These_DD2 signals_NN2 include_VV0 the_AT 'activity_NN1 level'_NN1 ,_, effectively_RR the_AT fraction_NN1 of_IO time_NNT1 the_AT person_NN1 speaks_VVZ ;_; their_APPGE 'engagement'_NN1 or_CC how_RGQ much_RR a_AT1 person_NN1 drives_VVZ the_AT conversation_NN1 ;_; and_CC 'mirroring'_NN1 ,_, which_DDQ occurs_VVZ when_RRQ one_MC1 participant_NN1 subconsciously_RR copies_VVZ another_DD1 's_VBZ prosody_NN1 and_CC gesture_NN1 ._. 
Pentland_NN1 is_VBZ well_JJ aware_JJ that_CST not_XX everyone_PN1 has_VHZ embraced_VVN this_DD1 vision_NN1 of_IO a_AT1 hidden_JJ ,_, pervasive_JJ 'second_NN1 channel'_NN1 in_II human_JJ affairs_NN2 ._. 
Although_CS the_AT idea_NN1 is_VBZ reasonably_RR well_RR accepted_VVN for_IF other_JJ animals_NN2 ,_, Pentland_NP1 says_VVZ that_CST the_AT reluctance_NN1 to_TO accept_VVI it_PPH1 in_II humans_NN2 lies_VVZ in_RR21 part_RR22 because_II21 of_II22 an_AT1 inherent_JJ bias_NN1 in_II how_RRQ science_NN1 looks_VVZ at_II human_JJ behaviour_NN1 versus_II that_DD1 of_IO other_JJ species_NN ._. 
"_" If_CS our_APPGE data_NN were_VBDR collected_VVN from_II ape_NN1 troops_NN2 ,_, "_" he_PPHS1 says_VVZ ,_, "_" and_CC we_PPIS2 altered_VVD the_AT semantically_RR loaded_JJ labels_NN2 a_RR21 bit_RR22 ,_, talked_VVN about_II 'forage'_NN1 instead_II21 of_II22 'work'_NN1 ,_, 'food_NN1 access_NN1 '_GE instead_II21 of_II22 'salary'_NN1 ,_, they_PPHS2 would_VM feel_VVI entirely_RR unsurprising_VVG ._. "_" 
Keep_VV0 it_PPH1 simple_JJ Behavioural_JJ science_NN1 ,_, as_CSA he_PPHS1 sees_VVZ it_PPH1 ,_, ought_VMK to_TO seek_VVI the_AT simplest_JJT explanation_NN1 for_IF human_JJ behaviour_NN1 first_MD ,_, looking_VVG to_II simple_JJ social_JJ signals_NN2 ,_, before_II constructing_VVG more_RGR complicated_JJ explanations_NN2 based_VVN on_II language_NN1 and_CC conscious_JJ reasoning_NN1 ._. 
Indeed_RR ,_, he_PPHS1 says_VVZ ,_, humans_NN2 lived_VVN in_II social_JJ groups_NN2 long_RR before_II language_NN1 evolved_VVD ,_, and_CC the_AT language_NN1 function_NN1 presumably_RR exists_VVZ on_II31 top_II32 of_II33 a_AT1 more_RGR archaic_JJ brain_NN1 system_NN1 for_IF non-linguistic_JJ social_JJ signalling_NN1 ._. 
The_AT behaviour_NN1 of_IO non-human_JJ primates_NN2 supports_VVZ this_DD1 idea_NN1 ,_, says_VVZ primatologist_NN1 Frans_NP1 de_NP1 Waal_NP1 of_IO Emory_JJ University_NN1 in_II Atlanta_NP1 ,_, Georgia_NP1 ._. 
Apes_NN2 ,_, chimpanzees_NN2 and_CC other_JJ primates_NN2 --_NN1 our_APPGE close_JJ evolutionary_JJ cousins_NN2 --_JJ lack_NN1 anything_PN1 like_II our_APPGE facility_NN1 for_IF language_NN1 ,_, yet_RR still_RR lead_VV0 sophisticated_JJ social_JJ lives_NN2 through_II displays_NN2 of_IO power_NN1 ,_, meaningful_JJ noises_NN2 and_CC facial_JJ expressions_NN2 ._. 
So_RR it_PPH1 is_VBZ "_" incredibly_RR naive_JJ "_" ,_, he_PPHS1 says_VVZ ,_, to_TO take_VVI conscious_JJ verbal_JJ communication_NN1 as_II the_AT primary_JJ way_NN1 that_CST people_NN respond_VV0 to_II each_PPX221 other_PPX222 ._. 
In_II the_AT future_NN1 ,_, Pentland_NP1 hopes_VVZ ,_, the_AT type_NN1 of_IO work_NN1 he_PPHS1 and_CC his_APPGE colleagues_NN2 have_VH0 been_VBN doing_VDG can_VM be_VBI expanded_VVN to_TO look_VVI in_II much_DA1 more_DAR detail_NN1 at_II human_JJ social_JJ interactions_NN2 in_II larger_JJR groups_NN2 ._. 
In_II an_AT1 ongoing_JJ experiment_NN1 at_II a_AT1 large_JJ German_JJ bank_NN1 ,_, for_REX21 example_REX22 ,_, they_PPHS2 are_VBR using_VVG their_APPGE sensors_NN2 to_TO focus_VVI on_II personal_JJ characteristics_NN2 that_CST are_VBR thought_VVN to_TO be_VBI linked_VVN to_II group_NN1 creativity_NN1 ._. 
According_II21 to_II22 recent_JJ work_NN1 on_II creativity_NN1 ,_, says_VVZ computer_NN1 scientist_NN1 Peter_NP1 Gloor_NP1 ,_, who_PNQS studies_VVZ collaborative_JJ innovation_NN1 networks_NN2 at_II MIT_NP1 's_GE Sloan_JJ School_NN1 of_IO Management_NN1 and_CC who_PNQS has_VHZ collaborated_VVN with_IW Pentland_NP1 ,_, creative_JJ people_NN tend_VV0 to_TO be_VBI more_RGR open_JJ and_CC agreeable_JJ ,_, and_CC less_RGR neurotic_JJ than_CSN others_NN2 ,_, and_CC they_PPHS2 tend_VV0 to_TO fluctuate_VVI more_RRR between_II extroversion_NN1 and_CC introversion_NN1 ._. 
These_DD2 traits_NN2 can_VM be_VBI assessed_VVN with_IW standard_JJ psychological_JJ questionnaires_NN2 ,_, but_CCB Pentland_NP1 's_GE group_NN1 thinks_VVZ the_AT sensors_NN2 can_VM do_VDI the_AT job_NN1 more_RGR effectively_RR ,_, giving_VVG managers_NN2 more_RGR reliable_JJ ways_NN2 to_TO identify_VVI patterns_NN2 of_IO human_JJ behaviour_NN1 ,_, and_CC thereby_RR to_TO assemble_VVI the_AT best_JJT possible_JJ creative_JJ teams_NN2 ._. 
"_" Collecting_VVG this_DD1 kind_NN1 of_IO spontaneous_JJ data_NN is_VBZ extremely_RR hard_JJ ._. 
But_CCB Alex_NP1 has_VHZ found_VVN a_AT1 way_NN1 to_TO do_VDI it_PPH1 ._. "_" 
Anna_NP1 Esposito_NP1 The_AT group_NN1 had_VHD 22_MC bank_NN1 employees_NN2 wear_VV0 data-collecting_JJ badges_NN2 for_IF one_MC1 month_NNT1 ,_, and_CC studied_VVD how_RRQ the_AT data_NN revealed_VVD personal_JJ characteristics_NN2 relevant_JJ to_II creativity_NN1 ._. 
For_REX21 example_REX22 ,_, from_II the_AT sensor_NN1 data_NN ,_, the_AT team_NN1 was_VBDZ able_JK to_TO construct_VVI a_AT1 'contribution_NN1 index'_NN1 for_IF each_DD1 person_NN1 that_CST measured_VVD ,_, roughly_RR speaking_VVG ,_, their_APPGE pro-activeness_NN1 in_II interpersonal_JJ interactions_NN2 ._. 
Someone_PN1 with_IW contribution_NN1 index_NN1 of_IO 1_MC1 looks_VVZ frequently_RR at_II other_JJ people_NN ._. 
Someone_PN1 with_IW an_AT1 index_NN1 of_IO -1_MC is_VBZ only_RR ever_RR looked_VVN at_II ,_, he_PPHS1 or_CC she_PPHS1 never_RR looks_VVZ another_DD1 actor_NN1 squarely_RR in_II the_AT face_NN1 ._. 
The_AT group_NN1 then_RT compared_VVD this_DD1 measure_NN1 with_IW descriptors_NN2 of_IO each_DD1 individual_NN1 as_CSA assessed_VVN in_II a_AT1 standard_JJ psychological_JJ test_NN1 ,_, judging_VVG variables_NN2 such_II21 as_II22 the_AT person_NN1 's_GE 'openness_NN1 '_GE towards_II fantasy_NN1 ,_, aesthetics_NN1 ,_, feelings_NN2 and_CC new_JJ ideas_NN2 ,_, or_CC their_APPGE 'extroversion'_NN1 ,_, spontaneous_JJ warmth_NN1 ,_, gregariousness_NN1 ,_, assertiveness_NN1 and_CC activity_NN1 ._. 
High_JJ values_NN2 of_IO the_AT index_NN1 ,_, they_PPHS2 found_VVD ,_, correlated_VVD very_RG strongly_RR with_IW the_AT psychological_JJ assessment_NN1 of_IO extroversion_NN1 ._. 
"_" The_AT more_RGR one_MC1 person_NN1 looks_VVZ at_II others_NN2 ,_, "_" says_VVZ Pentland_NP1 ,_, "_" the_AT more_DAR of_IO an_AT1 extrovert_NN1 they_PPHS2 seem_VV0 to_TO be_VBI ._. "_" 
Pentland_NN1 agrees_VVZ that_CST this_DD1 is_VBZ the_AT key_JJ challenge_NN1 ,_, but_CCB suggests_VVZ that_CST the_AT sensors_NN2 are_VBR orders_NN2 of_IO magnitude_NN1 more_RGR powerful_JJ than_CSN what_DDQ was_VBDZ used_VVN to_TO study_VVI human_JJ non-verbal_JJ interactions_NN2 in_II the_AT past_NN1 ._. 
"_" People_NN used_VVN to_II measure_NN1 mostly_RR short-term_JJ features_NN2 such_II21 as_II22 nods_NN2 or_CC smiles_NN2 ,_, whereas_CS we_PPIS2 can_VM make_VVI continuous_JJ ,_, quantitative_JJ measurements_NN2 over_RP much_RR longer_JJR times_NNT2 ,_, "_" he_PPHS1 says_VVZ ._. 
"_" We_PPIS2 can_VM also_RR examine_VVI precise_JJ details_NN2 of_IO frequency_NN1 and_CC amplitude_NN1 ,_, something_PN1 you_PPY ca_VM n't_XX do_VDI easily_RR without_IW automated_JJ tools_NN2 ._. "_" 
The_AT earlier_JJR work_NN1 ,_, he_PPHS1 suggests_VVZ ,_, was_VBDZ akin_JJ to_II trying_VVG to_TO understand_VVI written_JJ text_NN1 by_II looking_VVG just_RR at_II individual_JJ letters_NN2 ,_, ignoring_VVG words_NN2 or_CC sentences_NN2 ._. 
Moreover_RR ,_, he_PPHS1 suggests_VVZ ,_, almost_RR all_DB previous_JJ work_NN1 was_VBDZ done_VDN in_II the_AT lab_NN1 ,_, and_CC quantification_NN1 was_VBDZ all_RR21 but_RR22 impossible_JJ in_II anthropological-style_JJ field_NN1 work_NN1 ._. 
"_" In_II contrast_NN1 ,_, "_" Pentland_NP1 says_VVZ ,_, "_" we_PPIS2 are_VBR taking_VVG people_NN in_II real_JJ situations_NN2 ,_, in_II day-to-day_JJ work_NN1 settings_NN2 over_II months_NNT2 ,_, and_CC finding_VVG lots_PN of_IO reliable_JJ and_CC directly_RR predictive_JJ relationships_NN2 ._. "_" 
One_MC1 thing_NN1 he_PPHS1 does_VDZ admit_VVI is_VBZ that_CST experiments_NN2 of_IO this_DD1 sort_NN1 raise_VV0 the_AT spectre_NN1 of_IO Big-Brother-style_JJ intrusiveness_NN1 ._. 
Pentland_NN1 hopes_VVZ that_CST corporations_NN2 wo_VM n't_XX look_VVI at_II the_AT technology_NN1 as_CSA just_RR another_DD1 way_NN1 to_TO spy_VVI on_II employees_NN2 ,_, which_DDQ would_VM almost_RR guarantee_VVI resentment_NN1 and_CC loss_NN1 of_IO morale_NN1 ._. 
He_PPHS1 says_VVZ that_CST many_DA2 of_IO the_AT problems_NN2 could_VM be_VBI addressed_VVN if_CS the_AT companies_NN2 followed_VVD the_AT protocol_NN1 used_VVN in_II his_APPGE own_DA group_NN1 's_GE experiments_NN2 :_: the_AT data_NN were_VBDR first_MD stored_VVD locally_RR on_II each_DD1 subject_NN1 's_GE machine_NN1 ,_, so_CS21 that_CS22 at_II the_AT end_NN1 of_IO the_AT day_NNT1 ,_, he_PPHS1 or_CC she_PPHS1 could_VM review_VVI what_DDQ had_VHD been_VBN collected_VVN and_CC decide_VVI whether_CSW it_PPH1 would_VM be_VBI shared_VVN or_CC permanently_RR deleted_VVN ._. 
Pentland_NN1 also_RR says_VVZ that_CST employees_NN2 should_VM be_VBI given_VVN the_AT ability_NN1 to_TO turn_VVI off_II the_AT device_NN1 if_CS they_PPHS2 wish_VV0 ._. 
If_CS these_DD2 kinds_NN2 of_IO protections_NN2 are_VBR respected_VVN ,_, he_PPHS1 says_VVZ ,_, then_RT the_AT devices_NN2 could_VM lead_VVI to_II fundamentally_RR better_JJR science_NN1 ,_, including_II a_AT1 deeper_JJR knowledge_NN1 of_IO that_DD1 second_NNT1 channel_NN1 of_IO human_JJ social_JJ interaction_NN1 ._. 
"_" People_NN have_VH0 been_VBN studying_VVG this_DD1 kind_NN1 of_IO stuff_NN1 for_IF a_AT1 long_JJ time_NNT1 ,_, "_" says_VVZ Gazzaniga_NP1 ,_, "_" but_CCB Alex_NP1 is_VBZ raising_VVG it_PPH1 up_II21 to_II22 another_DD1 level_NN1 ._. 
It_PPH1 is_VBZ very_RG clever_JJ stuff_NN1 ._. "_" 
