Hidden Fees Exist Because They Work.

"Add-on fees are driving consumers crazy. From restaurants and hotels to concerts and food delivery, we are increasingly shown a low price online, only to click through and find fees that yield a much higher price at checkout.

Everyone says they hate these fees, but four experiments illustrate why "drip pricing," as it's called by researchers and regulators, is so effective at getting us to pay more. Note that this isn't about emotional blackmail, as with tipping: Drip pricing isn't negotiable. Nor does it explain why the cost of living might seem higher than the official data suggests: The consumer-price index reflects these fees and taxes.

The experiments explain why the fees proliferate. The conclusion: Consumers themselves are to blame. "Even when we know the fees are coming, we underestimate their magnitude," said Vicki Morwitz, a marketing professor at Columbia University.

The term drip pricing was popularized by a 2012 Federal Trade Commission conference. Its spread is associated with the proliferation of airline fees after 9/11. Yet an example of the phenomenon that long predates 2001 is stores' practice of listing goods without sales tax, which gets added at checkout.

Why not include the sales tax with the sticker price? A 2009 paper from economist Raj Chetty, then at UC Berkeley, and co-authors, showed consumers punish that sort of transparency.

A grocery store let the authors tag some products with the familiar pretax price and some with the total price including tax. For example, a hair brush's price tag showed $5.79 before tax, and beneath that $6.22 with the tax. Store managers predicted the transparency would be a disaster, and permitted the experiment for only three weeks and three product groups.

The managers were right. Sales volume dropped about 8% for products with price tags that included the tax than a control group without the tax.

This isn't because shoppers didn't know the tax rate or which items were taxable. In fact, 75% of shoppers surveyed knew the sales tax within a half percentage point, and most knew which goods were taxable. So the tax-inclusive price tag didn't give them new information; it was just that transparent reminders turned some people off.

For several years, Shannon Michelle White, a former Ph.D. student, and Abigail Sussman, her adviser at the University of Chicago, along with colleagues, have run a series of experiments asking participants to find the best deal for a range of purchases, such as wedding venues, prepaid cards, university tuition schedules, cellphone plans or home closing costs.

They can then choose between complex or simplified disclosures. For example, the complex disclosure for prepaid cards breaks down the final price into things such as "initial fees," "card acquisition fees," "card activation fees," "service fees" and "administration fees." The simplified disclosure combines all these fees as a single "initial fee."

If you realized this complexity tricks you into paying more for equivalent products, you're in the minority. Some 70% of people preferred the complex disclosure, said Sussman. They believe that it's more transparent and that they can calculate the total cost by themselves.

They're wrong. Even when participants are offered cash to identify the cheaper option, people botch the math (costly complexity!) and pick more expensive options.

In 2013, the website StubHub, which resells event tickets, attempted to do away with hidden fees, citing research about how hated they are. Its new "all-in pricing" prominently displayed the total ticket cost from the beginning of searches. The strategy failed to boost business or attract customers.

In 2015, shortly before abandoning all-in prices, StubHub did an experiment -- described several years later by economists who obtained the data -- where half of shoppers saw all-in pricing, and half saw the lower base price with taxes and fees only added at the end. The latter strategy boosted revenue 20%.

Shoppers didn't just buy more tickets. When they saw lower prices initially, they opted for better seats. By the end of the checkout process, they were committed.

"When people get to the end of the process, there's a variety of psychological reasons they're locked in," said Morwitz. "They overestimate the cost of starting over, they underestimate the benefits." Maybe they're just excited about the purchase, or reluctant to admit they could have made a mistake, she said.

Morwitz and co-authors Shelle Santana of Bentley University and Steven Dallas of Duke recruited people to book airlines and hotels for vacations, sometimes with rewards for finding the lowest rate.

People went for the low base prices, of course. After all the additional fees were dripped on, they had the option, not to mention incentive, to start over and look for a cheaper alternative. Here's the twist: Most opted not to, erroneously believing drip prices worked sort of like taxes, affecting the base price by a constant, fixed amount, the authors concluded in their 2020 paper for Marketing Science.

It isn't true. Drip prices can vary substantially between sellers, said Morwitz, and the lowest base price won't necessarily be the lowest final price.

"For consumers, it's hard," said Morwitz. "You have to look at the prices carefully, don't make any decisions until you see the total and be willing to restart the search."

Our instincts as consumers are unfortunately the opposite: We punish transparency, think we're clever enough to figure out any complexity, get sucked in by low offers then upgrade our ambitions, and conclude it's too much hassle to start over. When we shop that way, we have mostly ourselves to blame." [1]

1. U.S. News -- The Numbers: Hidden Fees Exist Because They Work. Zumbrun, Josh. 
Wall Street Journal, Eastern edition; New York, N.Y. [New York, N.Y]. 17 June 2023: A.2.

CRISPR vėžio tyrimo sėkmė atveria kelią individualizuotam gydymui

  "Nedidelis klinikinis tyrimas parodė, kad mokslininkai gali naudoti CRISPR genų redagavimą, kad pakeistų imunines ląsteles, kad jos atpažintų mutavusius baltymus, būdingus žmogaus navikams. Tada šias ląsteles galima saugiai paleisti organizme, kad būtų galima rasti ir sunaikinti jų taikinį.

 

     Tai pirmasis bandymas sujungti dvi karštas vėžio tyrimų sritis: genų redagavimą, kad būtų sukurtas individualus gydymas, ir imuninių ląstelių, vadinamų T ląstelėmis, inžinerija, siekiant geriau nukreipti į navikus. Metodas buvo išbandytas 16 žmonių, sergančių solidiniais navikais, įskaitant krūties ir storosios žarnos navikus.

 

     „Tai tikriausiai pati sudėtingiausia klinikoje kada nors bandyta terapija“, – sako tyrimo bendraautorius Antoni Ribas, vėžio tyrėjas ir Kalifornijos universiteto Los Andžele gydytojas. „Mes bandome sukurti armiją iš paties paciento T ląstelių."

 

     Rezultatai buvo paskelbti Nature ir pristatyti Vėžio imunoterapijos draugijos susitikime Bostone, Masačusetso valstijoje lapkričio 10 d.

 

     Pritaikytas gydymas

 

     Ribas ir jo kolegos pradėjo tirti DNR seką iš kraujo mėginių ir naviko biopsijų, siekdami ieškoti mutacijų, kurios randamos navike, bet ne kraujyje. Tai turėjo būti daroma kiekvienam tyrime dalyvaujančiam asmeniui. „Kiekvieno vėžio mutacijos yra skirtingos“, - sako Ribas. „Ir nors yra bendrų mutacijų, jos yra mažuma."

 

     Tada tyrėjai naudojo algoritmus, kad nuspėtų, kuri iš mutacijų gali sukelti atsaką iš T ląstelių – baltųjų kraujo kūnelių, kurie patruliuoja organizme, ieškant paklydusių ląstelių. „Jei [T ląstelės] mato kažką, kas atrodo nenormalu, jos tai nužudo“, – sako Stephanie Mandl, vyriausioji mokslinė pareigūnė PACT Pharma Pietų San Franciske, Kalifornijoje, ir pagrindinė tyrimo autorė. „Tačiau pacientų, kuriuos matome klinikoje, sergant vėžiu, imuninė sistema tam tikru momentu pralaimėjo kovą ir auglys išaugo."

 

     Atlikę daugybę analizių, kad patvirtintų savo išvadas, patvirtintų prognozes ir sukurtų baltymus, vadinamus T ląstelių receptoriais, galinčiais atpažinti naviko mutacijas, tyrėjai paėmė kraujo mėginius iš kiekvieno dalyvio ir naudojo CRISPR genomo redagavimą, kad įterptų šiuos receptorius koduojančius genus į paciento T ląsteles. Tada kiekvienas dalyvis turėjo vartoti vaistus, kad sumažintų jų gaminamų imuninių ląstelių skaičių, o po to sukurtos ląstelės buvo jam infuzuojamos.

 

     „Tai nepaprastai sudėtingas gamybos procesas“, – sako Josephas Fraietta, kuriantis T-ląstelių vėžio terapiją Pensilvanijos universitete Filadelfijoje. Kai kuriais atvejais visa procedūra truko ilgiau nei metus.

 

     Kiekvienas iš 16 dalyvių gavo inžinerines T ląsteles su iki trijų skirtingų taikinių. Vėliau buvo nustatyta, kad redaguotos ląstelės cirkuliuoja jų kraujyje ir jų buvo didesnėmis koncentracijomis šalia navikų, nei koncentracija neredaguotų ląstelių buvo prieš gydymą. Praėjus mėnesiui po gydymo penkiems dalyviams liga buvo stabili, o tai reiškia, kad jų navikai nebuvo išaugę. Tik du žmonės patyrė šalutinį poveikį, kurį tikriausiai lėmė redaguotų T ląstelių veikla.

 

     Nors gydymo veiksmingumas buvo mažas, mokslininkai naudojo palyginti mažas T ląstelių dozes, kad nustatytų metodo saugumą, sako Ribas. „Kitą kartą mums reikia tik smogti stipriau“, – sako jis.

 

     Tyrėjai plėtoja būdus, kaip paspartinti terapijos vystymąsi, inžinerijos būdu sukurtos ląstelės praleis mažiau laiko kultivuojamos už kūno ribų ir gali būti aktyvesnės, kai bus infuzuojamos. „Technologija bus geresnė ir geresnė“, – sako Fraietta.

 

     Tvirta pradžia

 

     Sukurtos T ląstelės, vadinamos CAR T ląstelėmis, buvo patvirtintos kai kurių kraujo ir limfos vėžio gydymui, tačiau kietieji navikai sukėlė ypatingą iššūkį. CAR T ląstelės yra veiksmingos tik prieš baltymus, kurie yra ekspresuojami naviko ląstelių paviršiuje. Tokių baltymų galima rasti daugelyje kraujo ir limfos vėžio atvejų, o tai reiškia, kad nereikia kurti naujų T ląstelių receptorių kiekvienam vėžiu sergančiam žmogui.

 

     Tačiau įprastų paviršiaus baltymų kietuose navikuose nerasta, sako Fraietta. Kietieji navikai sudaro fizines kliūtis T ląstelėms, kurios turi cirkuliuoti per kraują, nukeliauti iki naviko ir tada įsiskverbti į jį, kad nužudytų vėžines ląsteles. Auglio ląstelės taip pat kartais slopina imuninį atsaką, tiek išleisdamos imuninę sistemą slopinančius cheminius signalus, tiek išnaudodamos vietinį maistinių medžiagų tiekimą, kad paskatintų jų greitą augimą.

 

     „Aplinka aplink auglį yra kaip kanalizacija“, - sako Fraietta. „T ląstelės tampa mažiau funkcionalios, kai tik patenka į tą svetainę."

 

     Turėdami šį pradinį koncepcijos įrodymą, Mandl ir jos kolegos tikisi, kad T ląsteles pavyks ne tik atpažinti vėžio mutacijas, bet ir aktyviau veikti šalia naviko. Mandl sako, kad yra keletas galimų būdų sustiprinti T ląsteles, pavyzdžiui, pašalinant receptorius, kurie reaguoja į imunosupresinius signalus arba koreguojant jų metabolizmą, kad jie galėtų lengviau rasti energijos šaltinį naviko aplinkoje.

 

     Tokie sudėtingi projektai gali būti įmanomi dėl naujausių technologijų pažangos, naudojant CRISPR T ląstelėms redaguoti, sako Avery Posey, Pensilvanijos universitete tyrinėjantis ląstelių ir genų terapiją nuo vėžio. „Tai tapo neįtikėtinai efektyvu“, - sako jis. "Per ateinantį dešimtmetį pamatysime labai sudėtingas imuninių ląstelių inžinerijos priemones." [1]

1. Nature 611, 433-434 (2022)

CRISPR cancer trial success paves the way for personalized treatments

 

"A small clinical trial has shown that researchers can use CRISPR gene editing to alter immune cells so that they will recognize mutated proteins specific to a person’s tumours. Those cells can then be safely set loose in the body to find and destroy their target.

It is the first attempt to combine two hot areas in cancer research: gene editing to create personalized treatments, and engineering immune cells called T cells so as to better target tumours. The approach was tested in 16 people with solid tumours, including in the breast and colon.

“It is probably the most complicated therapy ever attempted in the clinic,” says study co-author Antoni Ribas, a cancer researcher and physician at the University of California, Los Angeles. “We’re trying to make an army out of a patient’s own T cells.”

The results were published in Nature¹ and presented at the Society for Immunotherapy of Cancer meeting in Boston, Massachusetts on 10 November.

Tailored treatments

Ribas and his colleagues began by sequencing DNA from blood samples and tumour biopsies, to look for mutations that are found in the tumour but not in the blood. This had to be done for each person in the trial. “The mutations are different in every cancer,” says Ribas. “And although there are some shared mutations, they are the minority.”

The researchers then used algorithms to predict which of the mutations were likely to be capable of provoking a response from T cells, a type of white blood cell that patrols the body looking for errant cells. “If [T cells] see something that looks not normal, they kill it,” says Stephanie Mandl, chief scientific officer at PACT Pharma in South San Francisco, California, and a lead author on the study. “But in the patients we see in the clinic with cancer, at some point the immune system kind of lost the battle and the tumour grew.”

After a series of analyses to confirm their findings, validate their predictions and design proteins called T-cell receptors that are capable of recognizing the tumour mutations, the researchers took blood samples from each participant and used CRISPR genome editing to insert the genes encoding these receptors into their T cells. Each participant then had to take medication to reduce the number of immune cells they produced, and the engineered cells were infused.

“This is a tremendously complicated manufacturing process,” says Joseph Fraietta, who designs T-cell cancer therapies at the University of Pennsylvania in Philadelphia. In some cases, the entire procedure took more than one year.

Each of the 16 participants received engineered T cells with up to three different targets. Afterwards, the edited cells were found circulating in their blood, and were present in higher concentrations near tumours than non-edited cells had been prior to the treatment. One month after treatment, five of the participants experienced stable disease, meaning that their tumours had not grown. Only two people experienced side effects that were probably due to the activity of the edited T cells.

Although the efficacy of the treatment was low, the researchers used relatively small doses of T cells to establish the safety of the approach, says Ribas. “We just need to hit it stronger the next time,” he says.

And as researchers develop ways to speed up the therapies’ development, the engineered cells will spend less time being cultured outside of the body and could be more active when they are infused. “The technology will get better and better,” says Fraietta.

A solid start

Engineered T cells — called CAR T cells — have been approved for the treatment of some blood and lymph cancers, but solid tumours have posed a particular challenge. CAR T cells are effective only against proteins that are expressed on the surface of tumour cells. Such proteins can be found across many blood and lymph cancers, which means there is no need to design new T-cell receptors for each person with cancer.

But common surface proteins have not been found in solid tumours, says Fraietta. And solid tumours provide physical barriers to T cells, which must circulate through the blood, travel to the tumour and then infiltrate it to kill the cancer cells. Tumour cells also sometimes suppress immune responses, both by releasing immune-suppressing chemical signals and by using up the local supply of nutrients to fuel their rapid growth.

“The environment around a tumour is like a sewer,” says Fraietta. “T cells are rendered less functional as soon as they hit the site.”

With this initial proof of concept in hand, Mandl and her colleagues hope to be able to engineer T cells not only to recognize cancer mutations, but also to be more active near the tumour. Mandl says there are several potential ways to toughen up T cells, for example by removing the receptors that respond to immunosuppressive signals, or by tweaking their metabolism so that they can more easily find an energy source in the tumour environment.

Such elaborate designs could be feasible thanks to recent technological advances in using CRISPR to edit T cells, says Avery Posey, who studies cell and gene therapies for cancer at the University of Pennsylvania. “It’s become incredibly efficient,” he says. “We’ll see very sophisticated means of engineering immune cells within the next decade.”" [1]

1. Nature 611, 433-434 (2022)

Dirbtinio intelekto startuoliai turi pinigų, bet jiems reikia daugiau duomenų

   „Generatyvinio dirbtinio intelekto (AI) pradedančios įmonės finansuojamos milijardais dolerių, tačiau jos jau gali patirti nesėkmę, jei negalės gauti reikiamų duomenų – ir tai nebus lengvas žygdarbis.

 

     Šiandien norint pasiekti sėkmę labai svarbu turėti tinkamus duomenis. Dabar, kai faktinių modelių kūrimas tapo šiek tiek preke, kurią rinka laiko lygiaverte arba beveik tokia, neatsižvelgiant į tai, kas ją pagamino, tikroji vertė yra duomenyse, sakė Paulas Tyma, „Bullpen Capital“ CTO.

 

     Remiantis „PitchBook“ duomenimis, rizikos finansavimas generatyvaus AI pradedančiose įmonėse išaugo nuo 4,8 mlrd. dolerių 2022 m. iki 12,7 mlrd. dolerių per pirmuosius penkis 2023 m. mėnesius. Dabar daugelis šių įmonių siekia sukurti daugiau nišinių AI modelių tokiose srityse kaip finansai ar sveikatos priežiūra, tačiau gauti prieigą prie mokymo duomenų rinkinių nebuvo lengva.

 

     Kai kurios AI pradedančiosios įmonės siekia partnerystės su didelėmis, daug duomenų turinčiomis įmonėmis. Pavyzdžiui, Marna Ricker, pasaulinė EY viceprezidentė mokesčių klausimais, teigė, kad įmonė turi generatyvaus dirbtinio intelekto pradedančiųjų įmonių, kurios kasdien į ją kreipiasi dėl daugybės sandorių duomenų. Tačiau EY pasaulinis vadovaujantis klientų aptarnavimo partneris Andy Baldwin teigė, kad nerimauja dėl to, kas nutiks EY duomenims, jei jie bus naudojami išoriniam modeliui apmokyti.

 

     "Kam tie duomenys priklauso? O kai mes mokome modelį, kokios yra mūsų prieigos prie to modelio teisės? O kaip kiti žmonės galės naudoti tą modelį?" - pasakė Baldwinas. „Duomenys yra dalis mūsų IP, kurį atsinešame.“" [1]

 

1. Business News: AI Startups Have Cash But Need More Data. Bousquette, Isabelle. 
Wall Street Journal, Eastern edition; New York, N.Y. [New York, N.Y]. 16 June 2023: B.5.