The decider

As a graduate student in the Department of Cognitive and Neural Systems at Boston University, I expected to use my research training for a career in brain modeling, robot design, even biomedical engineering. However, I recently stumbled upon a more unexpected application of my PhD skills: winning $500,000 on "Who Wants to Be a Millionaire."

Game shows have experienced a renaissance this past year, with NBC releasing two prime-time hits: "Deal or No Deal," a what's-behind-suitcase-number-two gambling show derived from Monty Hall's classic "Let's Make a Deal," and "1 vs 100," a quiz show pitting a single contestant against a mob of brainiacs and ordinary folk, taking its pedigree from "Millionaire."

ABC is about to launch the quiz show "Show Me the Money," hosted by William Shatner. Even the venerable "Millionaire" itself, reduced to daytime syndication after ABC overexposed its ratings-shattering Regis version, has witnessed a surge in interest since flirtatious and benevolent host Meredith Vieira was named co-anchor of the "Today" show.

I was drawn to "Millionaire" as a result of this renewed attention -- and some new ideas. My academic advisers Gail Carpenter and Stephen Grossberg investigate brain mechanisms essential for a profitable performance on "Millionaire": learning, memory, and (most importantly) decision-making. This summer, I tried to use their cognitive tricks to overcome the perils of the hot seat . (My appearance aired this Tuesday and Wednesday.)

In July, I rode the $15 Fung Wah bus to Manhattan and passed a multiple-choice audition test at the ABC commissary. Shortly afterward , I received the call to appear on "Millionaire." I prepared by analyzing the show's four lifelines: Ask-the-Audience, Phone-a-Friend, 50/50, and the new Switch-the-Question. An intelligent application of these is the difference between settling for the $25,000 safe haven or vaulting ahead to six-figure rewards.

The Ask-the-Audience allows a contestant to poll the studio audience's opinions. Conventional wisdom holds you should always trust the audience for questions under $25,000 and almost never for questions above.

Carpenter's cognitive research demonstrates that people's opinions tend to converge on certain responses: either the correct response or a response that seems correct. Of course, distinguishing between a correct response and a seemingly correct response is the very problem the "Millionaire" contestant is trying to solve. But her models suggest that if a contestant is able to determine where his own perspective fits within a population of opinions, it can be possible to correctly interpret Ask-the-Audience results.

I was presented with a textbook case of this on my $100,000 question: "In the famous scene where Marilyn Monroe stands atop a subway grate, what does she say as her skirt blows up?" Though I've certainly seen this clip, I don't believe I've ever heard the accompanying audio. I was sure the correct answer was not "Isn't it delicious?", which sounded too risqué, even for Monroe. "See anything you like?", which could have been uttered in a sarcastic or confrontational way, seemed the most reasonable choice. I unleashed the Ask-the-Audience: 40 percent thought it was "Isn't it delicious?", 40 percent thought "See anything you like?"

This was the exact scenario I had prepped for: I realized I was a member of the ignorant population who had chosen "See anything you like?", shunning "delicious . " Those spectators who had actually listened to that scene in "The Seven Year Itch" had selected the correct but counterintuitive response -- at least counterintuitive to those of us who hadn't heard the scene .

This entire analysis passed vividly through my mind -- but my brain flooded with neurohormones of fear and panic. I abandoned my hard-won faith in my graduate training. Instead I wasted the audience's gift and reached for my security blanket: the Phone-a-Friend lifeline.

I spent the most effort preparing for the Phone-a-Friend, a lifeline that grants contestants 30 seconds to repeat a question to a colleague. It's the only lifeline where a contestant has the possibility of being certain about the correct response -- if the answer is right in front of the friend, in a set of Google results.

But Googling a correct answer is no simple task. More often, questions resist concise search formulations: "Which of the following celebrities did not win an Emmy, an Oscar, a Tony, and a Grammy?"

Fortunately, Grossberg's cognitive neuroscience research focuses on brain-based technologies, including search algorithms that emulate the way the human cortex retrieves memories. I worked with fellow BU student Chaitanya Sai to develop Google- and Wikipedia-based software that analyzes each of the four possible answers. This analysis replicates the way brains simultaneously seek out memories along multiple pathways.

Of course, the most important component of our system was the human interface: I had to pose a query concisely and perfectly, while Sai had to enter the query flawlessly and interpret results swiftly. During rehearsals on hundreds of questions from previous shows, we answered them correctly 90 percent of the time . But we also learned that I would need the fortitude to recite a long and detailed string of search terms at a calm, steady pace that would permit Sai to accurately parse them.

It took me almost 20 seconds to enumerate the elaborate search string for the Monroe question, leaving Sai with only 10 seconds to run and interpret our search system. When the clock reached one second, the audience sighed with disappointment.

"Isn't it . . . ," Sai shouted as the timer buzzed and his voice was cut off. But it was time enough.

"B, 'Isn't it delicious?', final answer."

The 50/50 lifeline eliminates two wrong answers, leaving the correct answer and one incorrect answer. Deciding when and how to use the 50/50 lifeline requires contestants to accurately evaluate their own inner confidence. What if you have a feeling about an answer? What should you do with your hunch?

Though intuition sometimes serves us well, humans are generally very bad at evaluating probabilities. Our neural machinery has been shaped by millions of years of evolution to have certain innate biases when sizing up odds . Women tend to overestimate danger and men tend to overestimate their own competence. In fact, cognitive research has demonstrated very little correlation between our evaluation of the probability of an event and its frequency.

For example, many "Millionaire" contestants and viewers are convinced that on a 50/50 the producers usually ensure the two most plausible answers are left. It certainly seems that the 50/50 is biased against contestants. In fact, a statistical analysis reveals the two most plausible choices are left about one-third of the time: exactly what you would expect from a random process.

On my $500,000 question, "Which Beatle never appeared on Jerry Lewis' s Telethon?", I had no clue. I Switched-the-Question. My substitute question was almost as obscure: "When Bayer marketed heroin to consumers in the late 19th century, it was promoted as a remedy for what ailment?" I had an intuition it might be " persistent cough ." How could I objectively evaluate this hunch?

During my preparations, I reviewed every $500,000 and $1,000,000 question asked during the previous eight seasons, recording the times I felt a hunch and how often these guesses were accurate. I discovered my hunches were correct 50 percent of the time. This gave me objective odds for my "cough" hunch -- too low to merit gambling a quarter-million dollars.

I used my 50/50. "Persistent cough " remained, along with " stuffy head " -- which meant the odds my hunch was correct had risen to 75 percent . I remembered my promise to my wife that I wouldn't risk my $250,000 unless I was certain. But I decided to risk the quarter-million -- and a lifetime of marital strife.

"D, persistent cough, final answer."

And, just like that, I had reached the million-dollar question: "Which of the following four ships was not present at the Boston Tea Party? A. Eleanor, B. Dartmouth, C. Beaver, D. William."

Esoteric, but certainly not impossible: As a decade-long resident of Boston, surely the names of the ships must have passed through my awareness during some Freedom Trail walk or Duck Tours ride. I was pretty confident Dartmouth was at the Tea Party. I had a strong hunch Beaver was there, a weak hunch about Eleanor.

So did all those hunches about wrong answers equal one hunch about the right answer -- giving me a 50 percent shot at the million? Or was there some more complicated permutation of probabilities I needed to decipher? I felt my odds had to be better than 50 percent, but under the sweaty heat of the spotlights, the math melted in my brain like a jellyfish in the sun.

"I think the answer is William. . ."

I blinked. What would my advisers do?

". . . but I believe I'll walk with the money I've got. That's my final answer."

In the tranquility of my Waltham apartment, I calculated that my true odds had hovered near 75 percent. If I had managed to work that out in the hot seat , would I have chanced it? All I can say with certainty is that when the answer flashed on the screen, I felt like my heart had been clawed out of my chest.

"William."

See related article.